JP2008058956A - Speech reproduction device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech reproduction device capable of performing appropriate speed change in compliance with an inputted audio signal while achieving the target time. <P>SOLUTION: The speech reproduction device is provided with; a discrimination means which discriminates a speech section including a speech and a non-speech section not including the speech with respect to an audio signal; a speech content calculation means which calculates the speech content indicating the ratio of the speech section included in the audio signal based on the result discriminated by the discrimination means; a speed ratio calculation means which calculates respectively the speed ratio of the speech section and the non-speech section with respect to the reproduction speed preset in the audio signal based on the speech content in such a manner that the reproduction time of the audio signal becomes a prescribed reproduction time; and a speed change means which is inputted with the audio signal and respectively changes the reproduction speeds of the speech section and non-speech section included in the audio signal based on the speed ratio. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an audio reproduction device, and more particularly to an audio reproduction device that reproduces an audio signal by changing the reproduction speed.

  2. Description of the Related Art Conventionally, as an audio reproduction device that reproduces audio signals by changing the reproduction speed, an audio reproduction device that separately changes the reproduction speed of an audio section that includes sound and the reproduction speed of a non-audio section that does not include sound has been proposed. (For example, refer to Patent Document 1). Hereinafter, a conventional audio reproducing apparatus will be described with reference to FIG. FIG. 33 is a block diagram showing a configuration of a conventional audio reproducing apparatus.

In the conventional audio reproduction device shown in FIG. 33, the user sets a target time for the reproduction time of the entire audio signal. This target time is set to be shorter than the playback time when the entire audio signal is played at the same playback speed. The acoustic analysis unit 91 separates the input audio signal into a voice section and a non-voice section. For the audio signal in the audio section sandwiched between the non-audio sections of a certain time length or longer, the speed conversion unit 92 has a beginning portion that is slower than the predetermined reproduction speed and gradually increases toward the end. Speed conversion is performed to return to Here, due to the speed conversion processing in the speech speed conversion unit 92, there is a problem that the reproduction time of the voice section becomes long, and as a result, the reproduction time of the entire audio signal is delayed with respect to the target time. Therefore, the non-speech interval length control unit 93 refers to the delay time information output from the speed conversion unit 92 and performs processing for shortening the delay time for the non-speech interval. Specifically, the non-speech section length control unit 93 performs a process of deleting or compressing the non-speech section to shorten the delay time. The audio signal of the voice section that has been speed-converted by the speed conversion unit 92 and the audio signal of the non-speech section processed by the non-speech section length control unit 93 are synthesized by the synthesis unit 94 and output from the synthesis unit 94.
JP 2001-222300 A (page 1-2, FIG. 1)

  Here, the ratio of the voice sections included in the input audio signal differs depending on the input audio signal. However, in the conventional audio reproduction device, regardless of the ratio of the audio section, the speed conversion process is uniformly performed for the audio section, and deletion or non-audio section for achieving the target time is performed. Compression is in progress. Therefore, for example, when the input audio signal is a signal including a lot of speech sections, the speed conversion process is uniformly performed on the speech sections, so that the delay time generated in the speech speed conversion unit 92 becomes long. . If the delay time is increased, the amount of deletion and compression of the non-voice interval is increased, and the loss of information becomes large and the reproduction becomes difficult to hear. As described above, in the conventional audio reproduction device, it is impossible to perform an appropriate speed conversion according to the input audio signal while achieving the target time.

  SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide an audio reproduction device capable of performing appropriate speed conversion according to an input audio signal while achieving a target time.

  1st invention is the audio | voice reproduction apparatus which changes the reproduction | regeneration speed of the audio signal of the content input, and reproduces | regenerates by predetermined reproduction time, Comprising: With respect to an audio signal, the audio area which contains an audio | voice, and an audio | voice are not included. A discriminating unit for discriminating a non-speech section; a voice content rate calculating unit for calculating a voice content rate indicating a ratio of a voice segment included in the audio signal based on a discrimination result discriminated by the discriminating unit; A speed ratio calculating means for calculating a speed ratio between a voice section and a non-speech section with respect to a playback speed set in advance in the audio signal based on the voice content so that the playback time becomes a predetermined playback time; Speed conversion means for converting the playback speeds of the speech and non-speech sections included in the audio signal based on the speed ratio. .

  According to a second aspect, in the first aspect, the speed ratio calculation means includes a companding ratio indicating a ratio of companding a reproduction time reproduced at a reproduction speed set in advance to the audio signal to a predetermined reproduction time. Using the correspondence information indicating correspondence between the voice content rate, the voice average speed ratio indicating the average speed ratio of the voice section, and the non-voice average speed ratio calculating method of the average speed ratio of the non-voice section, Speed ratio condition setting means for calculating the ratio and the non-speech average speed ratio and setting the ratio as a speed ratio condition, and determining the speed ratio in each section into which the speech section is subdivided as a speed ratio based on the speech average speed ratio And a speed ratio determining means for determining a speed ratio based on a non-speech average speed ratio and calculating a speed ratio between the speech section and the non-speech section.

  In a third aspect based on the second aspect, the sound reproducing device further comprises a sound section length calculating means for calculating the time from the start time to the end time of each sound section determined by the determining means as a sound section length. The speed ratio condition setting means further sets an end speed ratio at the end time of the voice section according to the voice average speed ratio as a speed ratio condition, and the speed ratio determination means is configured to determine each voice section determined by the determination means. On the other hand, the speed ratio in each section of the voice section is determined based on the voice average speed ratio, the voice section length, and the end speed ratio.

  In a fourth aspect based on the second aspect, the speed ratio determination unit obtains the time elapsed from the start time of the speech segment by the speech segment length for each speech segment determined by the determination unit. The speed ratio in each section of the voice section is determined according to the elapsed ratio.

  In a fifth aspect based on the second aspect, the speed ratio determining means determines the speed ratio in each section of the voice section so that the playback speed increases as time elapses from the start time of the voice section.

  In a sixth aspect based on the second aspect, the speed ratio condition setting means is correspondence information including at least one type of calculation method of the voice average speed ratio and the non-voice average speed ratio, and is configured by an audio signal. Using the correspondence information indicating different correspondence depending on the type of content, the voice average speed ratio and the non-voice average speed ratio are calculated.

  In a seventh aspect based on the second aspect, the speed ratio condition setting means creates correspondence information so that the voice average speed ratio and the non-voice average speed ratio are within a range designated by the user, and the correspondence The voice average speed ratio and the non-voice average speed ratio are calculated using the information.

  In an eighth aspect based on the second aspect, the correspondence information indicates a correspondence in which the voice content rate differs from the calculation method of the voice average speed ratio and the non-voice average speed ratio according to the magnitude of the voice content rate. Information.

  In a ninth aspect based on the second aspect, the speed ratio condition setting means is correspondence information including at least one kind of calculation method of the voice average speed ratio and the non-voice average speed ratio, and corresponds to a user's purpose of use. Using the correspondence information indicating different correspondences, the voice average speed ratio and the non-voice average speed ratio are calculated.

  In a tenth aspect based on the second aspect, the correspondence information indicates a correspondence in which the companding ratio differs from the calculation method of the voice average speed ratio and the non-voice average speed ratio according to the magnitude of the companding ratio. Information.

  According to an eleventh aspect of the invention, there is provided storage means for preliminarily storing the audio signal constituting the entire content and the discrimination result discriminated by the discrimination means for the audio signal constituting the entire content in the first invention. The voice content rate calculating unit calculates a voice content rate indicating a ratio of the voice sections included in the audio signal constituting the entire content based on the determination result stored in advance in the storage unit.

  In a twelfth aspect based on the first aspect, the voice content rate calculating unit sequentially calculates the voice content rate used when the speed ratio calculating unit calculates based on the determination result previously determined by the determination unit. To do.

  In a thirteenth aspect based on the twelfth aspect, the voice content rate calculating means calculates the speed ratio based on the determination result determined in the past for the first predetermined time from when the speed ratio calculating means calculates. The audio content rate used when the means calculates is sequentially calculated every second predetermined time which is a time equal to or less than the first predetermined time, and the audio reproducing apparatus has the data amount input to the speed converting means and the speed converting means Based on the amount of data output from the audio signal and a companding ratio indicating a ratio of companding the reproduction time reproduced at the reproduction speed set in advance to the audio signal to the predetermined reproduction time. The drawing apparatus further includes a drawing ratio calculating unit that sequentially calculates the drawing ratio, and the speed ratio calculating unit includes a drawing ratio for each second predetermined time, a voice content rate for each second predetermined time, and an average of the voice section. Voice average speed ratio indicating speed ratio and flatness of non-voice section Using the correspondence information indicating the correspondence with the calculation method of the non-voice average speed ratio indicating the speed ratio, the voice average speed ratio and the non-voice average speed ratio are calculated, respectively, and the calculated voice average speed ratio and the non-voice average speed ratio are calculated. Speed ratio condition setting means for sequentially setting the ratio as a speed ratio condition every second predetermined time, and each segment in which the speech segment is subdivided with respect to the speech segment and the non-speech segment included in the second predetermined time Is determined to be a speed ratio based on the voice average speed ratio, and a speed ratio in each section in which the non-voice sections are subdivided is determined to be a speed ratio based on the non-voice average speed ratio, for a second predetermined time. Speed ratio determining means for calculating the speed ratio of the voice section and the non-voice section based on the speed ratio condition sequentially set for each, and the speed conversion means includes the voice section and the non-voice section included in the audio signal. Playback Converting each based degrees on the speed ratio of the calculated speech section and the non-speech section in the speed ratio determining means.

  In a fourteenth aspect based on the thirteenth aspect, the sound reproducing device is a statistic for suppressing a change at every second predetermined time indicated by the speed ratio in each section of the sound section determined by the speed ratio determining means. Statistic calculating means for calculating a quantity, and the speed ratio condition setting means is configured to calculate the voice average speed based on the statistic, the voice content rate for each second predetermined time, and the companding ratio for each second predetermined time. And the end speed ratio of the voice section based on the average voice speed, and sequentially set the calculated voice average speed and end speed ratio as speed ratio conditions every second predetermined time.

  In a fifteenth aspect based on the fourteenth aspect, the statistic calculation means is configured so that the speed ratio calculation means starts from the content start time based on the determination result from the content start time to the time when the speed ratio calculation means calculates. The voice content rate indicating the ratio of the voice sections included up to the time of calculation is calculated as a statistic.

  In a sixteenth aspect based on the thirteenth aspect, the sound reproducing device uses the start time to the end time of the sound section used when the speed ratio calculating means calculates based on the determination result determined in the past by the determining means. Voice section length calculating means for sequentially calculating the voice section length that is the time until the speed ratio condition setting means further uses the end speed ratio at the end time of the voice section according to the voice average speed ratio as a speed ratio condition. The speed ratio determining means sequentially sets every second predetermined time, and for each voice section discriminated by the discriminating means, each speed section is based on the voice average speed ratio, the voice section length, and the end speed ratio. Determine the speed ratio in the section.

  In a seventeenth aspect based on the sixteenth aspect, the voice section length used when the speed ratio calculating means calculates is the voice section length calculated from the start and end times of the voice section determined in the past by the determining means. Of these, it is calculated based only on the voice segment length that is greater than or equal to the predetermined segment length.

  In an eighteenth aspect based on the sixteenth aspect, the voice section length used when the speed ratio calculating means calculates is the voice section length calculated from the start and end times of the voice sections determined in the past by the determining means. It is calculated based on the maximum value of.

  According to a nineteenth aspect of the present invention, in the first aspect of the invention, further includes an accumulation unit that accumulates in advance the audio signal for a predetermined time and the determination result determined by the determination unit for the audio signal for the predetermined time, The voice content rate calculating unit calculates a voice content rate indicating a ratio of a voice section included in the audio signal for a predetermined time based on the determination result stored in advance in the storage unit.

  In a twentieth aspect based on the first aspect, the determining unit determines, for the audio signal, a specific event section including a specific event sound, a voice section other than the specific event section, and a non-voice section, The audio reproduction device further includes a specific event content rate calculating unit that calculates a specific event content rate indicating a ratio of a specific event section included in the audio signal based on the determination result determined by the determining unit, and a speed ratio calculating unit Calculates the speed ratio of the specific event section to the playback speed preset in the audio signal based on the specific event content rate, and sets the audio signal in advance so that the playback time of the audio signal becomes a predetermined playback time. Based on the audio content rate, the speed ratio of the audio and non-audio intervals other than the specific event interval to the playback speed It was calculated, the rate conversion means converts on the basis of the playback speed of a particular event section and the specific events other than segment speech segment and the non-speech section included in the audio signal to the speed ratio.

  A twenty-first aspect of the present invention is an audio reproduction method for reproducing at a predetermined reproduction time by changing the reproduction speed of the audio signal of the input content. The audio signal includes an audio section including audio and no audio. A discriminating step for discriminating a non-speech section; a voice content rate calculating step for calculating a voice content rate indicating a ratio of a voice segment included in the audio signal based on the discrimination result discriminated in the discriminating step; A speed ratio calculating step for calculating a speed ratio between a voice section and a non-voice section with respect to a playback speed set in advance in the audio signal based on the voice content so that the playback time becomes a predetermined playback time; As input, the speed at which the playback speed of the speech and non-speech sections included in the audio signal is converted based on the speed ratio. And a conversion step.

  A twenty-second aspect of the invention is a program for causing a computer of an audio reproduction device to reproduce at a predetermined reproduction time by changing the reproduction speed of an audio signal of input content, and includes audio for the audio signal. A speech step and a non-speech segment that does not include speech, and a speech inclusion that calculates a speech content ratio that indicates a ratio of speech segments included in the audio signal based on the discrimination result determined in the discrimination step A rate calculating step, and a speed for calculating a speed ratio of a voice section and a non-voice section with respect to a playback speed set in advance in the audio signal based on a voice content rate so that the playback time of the audio signal becomes a predetermined playback time. A ratio calculation step and an audio signal as an input, and a speech segment and a non-speech segment included in the audio signal are reproduced. A speed conversion step of converting each based on the speed of the speed ratio, a program executed by a computer.

  A twenty-third invention is a computer-readable recording medium on which the program of the twenty-second invention is recorded.

  A twenty-fourth aspect of the invention is an integrated circuit that plays back a predetermined playback time by changing the playback speed of the audio signal of the input content. The audio circuit includes a speech section that includes speech and a non-speech that does not include speech. A discriminating unit for discriminating a voice section; a voice content rate calculating unit for calculating a voice content rate indicating a ratio of a voice segment included in the audio signal based on the discrimination result discriminated by the discriminating unit; and reproduction of the audio signal Speed ratio calculating means for calculating the speed ratio of the voice section and the non-voice section with respect to the playback speed set in advance in the audio signal based on the voice content rate, and the audio signal so that the time becomes a predetermined playback time And a speed conversion means for converting the playback speed of the voice section and the non-voice section included in the audio signal based on the speed ratio, respectively.

  According to the first aspect, by calculating the audio content rate of the input audio signal, the audio content rate of the input audio signal is calculated based on the audio content rate of the input audio signal while achieving a predetermined reproduction time. In addition, it is possible to calculate the speed ratio between the optimum voice segment and the non-speech segment. In other words, by calculating the voice content rate, the ratio of the voice sections included in the input audio signal can be known, and the optimum speed for achieving the predetermined playback time for both the voice sections and the non-voice sections. The ratio can be calculated. This makes it possible to realize easy-to-listen playback with reduced discontinuity of playback content and discomfort due to lack of information, regardless of what audio signal is input.

  According to the second aspect of the invention, the voice average speed ratio and the non-voice average speed ratio are calculated based on the companding ratio indicating the ratio of companding during a predetermined reproduction time and the voice content rate, respectively, and the voice interval and the non-voice speed are calculated. By determining the speed ratio in each section of the voice section, the optimum speed ratio of the voice section and the non-voice section is calculated based on the voice content of the input audio signal while achieving a predetermined playback time. can do.

  According to the third aspect, the speed ratio at the end time of the voice section is terminated by determining the speed ratio in each section of the voice section based on the voice average speed ratio, the voice section length, and the end speed ratio. While maintaining the speed ratio constant, it is possible to determine a speed ratio suitable for the speech section length from the beginning of the sentence to the end of the sentence, and to reduce difficulty in hearing and unnaturalness due to high-speed playback at the end of the speech section.

  According to the fourth aspect, by determining the speed ratio in each section of the speech section according to the elapsed rate, the speed in each section of the speech section can be determined using a simple function regardless of the length of the speech section length. The ratio can be determined.

  According to the fifth aspect of the invention, since the playback speed of the beginning portion of the voice section is relatively slow compared to other portions, it is possible to prevent the understanding level of the playback content from being lowered due to missed listening to the beginning portion. it can.

  According to the sixth aspect, it is possible to calculate different voice average speed ratios and non-voice average speed ratios depending on the type of content, and the speed ratio in each section of the voice section and the non-voice section is determined according to the content. High accuracy can be achieved.

  According to the seventh aspect, the speed ratio in each of the speech section and the non-speech section becomes a speed ratio based on the speech average speed ratio and the non-speech average speed ratio within the range specified by the user, and the user's listening ability Speed conversion processing according to the user's preference.

  According to the eighth aspect of the invention, the optimal speed ratio between the voice segment and the non-speech segment is calculated based on the voice content rate of the input audio signal while achieving a companding ratio indicating a predetermined reproduction time. be able to.

  According to the ninth aspect, it is possible to change the calculation method of the voice average speed ratio and the non-voice average speed ratio depending on the purpose of use of the user. Various controls relating to the output time of the audio signal including insertion and pause can be handled. As a result, it is not necessary to create the content separately for viewing, overview comprehension, language learning, transcription, etc., and the same content can be used for various purposes.

  According to the tenth aspect of the present invention, the correspondence between the voice average speed ratio and the non-voice average speed ratio calculation method differs depending on the magnitude of the companding ratio, for example, when the companding ratio is low, In addition, when the companding ratio is high, it is possible to determine a speed ratio according to the user's purpose for learning purposes. As a result, the user can use the device without being aware of the proper use of the device according to the purpose, and can perform a speed conversion process in accordance with the viewing request of the user.

  According to the eleventh aspect of the present invention, by calculating the audio content rate for the entire content, it is possible to calculate the speed ratio between the voice segment and the non-speech segment with high accuracy.

  According to the twelfth aspect, since processing is possible without providing storage means, it is not necessary to wait for an audio signal to be stored in the storage means, and speed conversion processing can be performed in real time.

  According to the thirteenth aspect, the voice content rate for the first predetermined time is reflected in the second predetermined time, and the speed of the voice interval and the non-voice interval that immediately reflects the change in the voice content rate. The ratio can be calculated. Further, even if an error occurs in the reproduction time by calculating the audio content rate from the first predetermined time, the speed conversion process is performed using the companding ratio for each second predetermined time. The error can be eliminated in the speed conversion process from now on.

  According to the fourteenth aspect, by using the statistic for suppressing the change at the second predetermined time indicated by the speed ratio in each section of the voice section, the voice content ratio for the first predetermined time is used. Even when becomes locally high, it is possible to prevent an excessive increase in the speed ratio in each section of the speech section. As a result, it is possible to perform speed conversion processing corresponding to various contents having different audio contents.

  According to the fifteenth aspect of the present invention, the voice content rate for the first predetermined time is locally determined by using, as a statistic, the voice content rate that has a tendency to change over time from the voice content rate for the first predetermined time. Even if the frequency fluctuates automatically, the fluctuation is suppressed, and an easy-to-listen speed conversion process becomes possible.

  According to the sixteenth aspect of the present invention, since the speech section length is sequentially calculated, an appropriate speed conversion process can be performed on the speech section without knowing the end time of the speech section, and more accurate High real-time processing can be realized.

  According to the seventeenth aspect of the present invention, the calculation is performed based only on the voice section length that is equal to or longer than the predetermined section length, so that the average of the sum excluding the “yes” and “yeah” and the filler such as “e” is removed. It is possible to calculate a long speech interval length.

  According to the eighteenth aspect of the invention, the voice section length used when the speed ratio calculating means calculates is calculated based on the maximum value of the past voice section length, so that the end speed ratio is converted at the end time of the voice section. The ratio is reduced, the average speed ratio of the voice section is further reduced, and an easy-to-listen speed conversion process can be realized.

  According to the nineteenth aspect, speed conversion can be performed in units of audio signals for a predetermined time. As a result, even during content recording, speed conversion processing can be performed without waiting for the end of the entire recording. Further, by accumulating the discrimination results of the voice section and the non-speech section in the storage means, it is possible to calculate the actual value of the voice content rate, and to perform the speed conversion with a more optimal speed ratio.

  According to the twentieth invention, by calculating the specific event content rate and calculating the speed ratio of the specific event section, for example, when the specific event sound is music, the music section with respect to an audio signal such as a music program Can be reproduced at a slower reproduction speed than the voice and non-voice segments. As a result, it is possible to perform speed conversion processing with emphasis on music.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(First embodiment)
First, an audio reproducing apparatus according to the first embodiment of the present invention will be described with reference to FIG. FIG. 1 is a block diagram illustrating a configuration example of an audio reproduction device according to the first embodiment. In FIG. 1, the audio playback apparatus includes an audio non-audio discriminating unit 11, an accumulation unit 12, an audio content rate calculating unit 13, a speed ratio condition setting unit 14, an audio section length calculating unit 15, a speed ratio determining unit 16, and a speed The conversion unit 17 is configured. In the present embodiment, an audio reproduction apparatus that accumulates audio signals to be speed-converted in the accumulation unit 12 in units of contents in advance and performs a reproduction process using the accumulated audio signals with different reproduction speeds will be described. Further, in the following description, a section including voice is referred to as a voice section. A section other than the voice section, that is, a section that does not include voice is defined as a non-voice section.

  The voice / non-voice discrimination unit 11 receives an audio signal and discriminates between a voice zone and a non-voice zone. Further, the voice non-voice discrimination unit 11 outputs the start / end time (start time and end time) of the voice section together with the discrimination result. The input audio signal is an audio signal recorded on a CD, DVD, memory, hard disk, or the like. Note that the audio signal may be an audio signal distributed via a communication line such as the Internet or an audio signal received by broadcasting. The audio signal may be generated on the spot such as voice synthesis, recorded with a microphone, or output through a communication device such as a telephone. Here, as a method of discriminating between the voice segment and the non-speech segment, for example, a method of calculating the power of the audio signal and discriminating with a threshold value can be cited. In addition, as described in, for example, “Speech and Music Segmentation Using Cepstrum Flux” <SP2000-1, Takayuki Uchida, Masami Yamashita, Masahide Sugiyama, Shingaku Technical Report, SP2000-17> There is also a method of measuring and discriminating. In the method of measuring the degree of change of the cepstrum, it is possible to discriminate even a voice on which BGM is superimposed.

  The storage unit 12 includes a readable / writable recording medium such as a hard disk, a DVD, or a memory medium (for example, an SD card). The storage unit 12 stores the same audio signal that is input to the voice / non-voice discrimination unit 11 in units of contents. The storage unit 12 stores the determination result and the start / end time of the voice section output from the voice / non-voice discrimination unit 11. Here, for example, consider the case of recording a TV broadcast. In this case, the audio signal and the video signal constituting the TV broadcast are stored in the storage unit 12. Along with this accumulation, a discrimination process is performed in the voice non-voice discrimination unit 11, and the discrimination result and the start / end time of the voice section are accumulated in the accumulation unit 12. In the storage unit 12, an audio signal, a video signal, a discrimination result, and start / end times of an audio section are associated with each content and stored. The format of the audio signal and the video signal may be any format.

  The audio content rate calculation unit 13 calculates an audio content rate indicating the ratio of audio intervals included in the audio signal of the content. Specifically, the audio content rate calculation unit 13 calculates the audio content rate for each content stored in the storage unit 12 using the corresponding determination result and the start / end time of the audio section. The voice content rate is obtained by dividing the sum of the voice section lengths included in the audio signal for a predetermined time by the predetermined time. In the present embodiment, the audio content rate is obtained by dividing the sum of the audio section lengths included in the audio signal of the content by the content length. Here, the content means an entire program that performs speed conversion. Therefore, the content length is usually equal to the program length, and is often 30 minutes or 1 hour. In addition, when the user designates a part of the program as a speed conversion target, the part may be the content.

  The audio content varies depending on the content. For example, as shown in FIG. 2, when content is viewed by genre, it can be seen that the audio content varies depending on the genre. FIG. 2 is a diagram showing audio content rates by genre. In FIG. 2, the horizontal axis indicates the genre, and the vertical axis indicates the audio content rate. In addition, the audio content shown in FIG. 2 is obtained by extracting the top six programs in the audience rating by genre from the programs broadcast in the same week, and summing up the audio content for each extracted program by genre. Averaged. The audio content of news is about 60%, which is the highest value among the five genres. The audio content of sports and music is about 40%, which is nearly 20% wider than news. Even in the same genre, there is some variation in the audio content as shown in FIG. FIG. 3 is a diagram showing the average and standard deviation of the audio content of each genre. In drama and animation, the standard deviation is 16.2, which is higher than other genres. Thus, the audio content varies depending on the content.

  Therefore, in the conventional technology that does not consider the voice content rate, as described above, the delay time from the target time becomes long in a news program with a high voice content rate. As a result, there has been a problem that in order to eliminate the delay time, the voice section is partially reproduced at high speed or deleted, and the reproduced audio signal becomes difficult to hear. On the other hand, in this embodiment, the audio content rate of content is calculated. Accordingly, it is possible to calculate the optimum voice / non-voice interval speed ratio according to the content without delay from the target time, and it is possible to realize reproduction that is easy to hear without partial bias. A method for calculating the speed ratio using the voice content will be described in detail later.

  The speed ratio condition setting unit 14 receives the voice content rate and the target companding ratio, calculates the average speed ratio of the voice section, the average speed ratio of the non-voice section, and the terminal speed ratio of the voice section, and calculates these speed ratios. Set as a condition.

  The companding ratio is obtained by dividing the reproduction time length after the speed conversion process by the reproduction time length before the speed conversion process. In the normal speed reproduction, the companding ratio is 1. In the double speed reproduction, the draw ratio is 0.5. When the companding ratio takes a value from 0 to 1, the playback time length is compressed and played back at a speed faster than the normal speed. When the companding ratio takes a value greater than 1, the reproduction time length is extended and reproduction is performed at a speed slower than the normal speed. The target companding ratio indicates how much the playback time length of the content to be speed-converted is compressed or expanded. The target companding ratio takes a value from 0 to 1 in the case of compression, and takes a value of 1 or more in the case of expansion. The target companding ratio may be input by the user or may be set in advance in the apparatus. Further, the user may not directly input the target companding ratio. In this case, a target time for content reproduction is input. When the user inputs the target time, the target companding ratio can be obtained by dividing the target time by the reproduction time length before the speed conversion process. The speed ratio indicates the ratio of the speed to the normal speed. The speed ratio is represented by the reciprocal of the drawing ratio. Further, the termination speed ratio of the voice section means the speed ratio at the termination time of the voice section.

  Next, a method for calculating the average speed ratio between voice and non-voice sections will be described. The speed ratio condition setting unit 14 calculates an average speed ratio using a preset speed ratio calculation distribution. The speed ratio calculation distribution is a distribution indicating in which calculation pattern the average speed ratio is calculated according to the voice content ratio and the target companding ratio. In other words, the speed ratio calculation distribution is correspondence information indicating the correspondence between the voice content rate, the target companding ratio, and the method of calculating the average speed ratio between the voice and non-voice sections.

なお、Ｓは音声含有率、Ｖｍ１は音声区間の平均速度比、Ｖｍ２は非音声区間の平均速度比、Ｅは目標圧伸比を示す。算出パターンとしては、図４に示すように５種類の算出パターンａ〜ｅが考えられる。図４は、５種類の算出パターンを示した図である。算出パターンａ〜ｅの条件は、以下のようになる。
ａ：非音声区間の平均速度比Ｖｍ２＝Ａｎ（固定値）として、与えられる音声含有率Ｓと目標圧伸比Ｅから式（１）を満たすように、音声区間の平均速度比Ｖｍ１を算出する。但し、Ｖｍ１≦Ａｎを算出条件とする。
ｂ：音声区間の平均速度比Ｖｍ１＝Ｂｓ（固定値）として、与えられる音声含有率Ｓと目標圧伸比Ｅから式（１）を満たすように、非音声区間の平均速度比Ｖｍ２を算出する。但し、Ｖｍ２≧Ｂｓを算出条件とする。
ｃ：音声及び非音声区間の平均速度比をＶｍ１＝Ｖｍ２として、与えられる音声含有率Ｓと目標圧伸比Ｅから式（１）を満たすように、音声及び非音声区間の平均速度比Ｖｍ１及びＶｍ２を算出する。
ｄ：非音声区間の平均速度比Ｖｍ２＝Ｄｎ（固定値）として、与えられる音声含有率Ｓと目標圧伸比Ｅから式（１）を満たすように、音声区間の平均速度比Ｖｍ１を算出する。但し、Ｖｍ１≧Ｄｎを算出条件とする。
ｅ：音声区間の平均速度比Ｖｍ１＝Ｅｓ（固定値）として、与えられる音声含有率Ｓと目標圧伸比Ｅから式（１）を満たすように、非音声区間の平均速度比Ｖｍ２を算出する。但し、Ｖｍ２≦Ｅｓを算出条件とする。 Hereinafter, the calculation pattern will be described. The average speed ratio between the voice and non-voice sections is calculated so as to achieve the target companding ratio. Specifically, it is calculated so as to satisfy Expression (1).

S represents the voice content rate, Vm1 represents the average speed ratio of the voice section, Vm2 represents the average speed ratio of the non-voice section, and E represents the target companding ratio. As the calculation patterns, there are five types of calculation patterns a to e as shown in FIG. FIG. 4 is a diagram showing five types of calculation patterns. The conditions of the calculation patterns a to e are as follows.
a: As the average speed ratio Vm2 = An (fixed value) in the non-speech section, the average speed ratio Vm1 in the speech section is calculated from the given speech content S and the target companding ratio E so as to satisfy the expression (1). . However, Vm1 ≦ An is a calculation condition.
b: As the average speed ratio Vm1 = Bs (fixed value) of the speech section, the average speed ratio Vm2 of the non-speech section is calculated from the given speech content S and the target companding ratio E so as to satisfy the formula (1). . However, Vm2 ≧ Bs is a calculation condition.
c: The average speed ratio Vm1 between the voice and the non-speech section is set so that the average speed ratio between the voice and the non-speech section is Vm1 = Vm2 Vm2 is calculated.
d: As the average speed ratio Vm2 = Dn (fixed value) of the non-voice section, the average speed ratio Vm1 of the voice section is calculated from the given voice content S and the target companding ratio E so as to satisfy the equation (1). . However, Vm1 ≧ Dn is a calculation condition.
e: The average speed ratio Vm2 of the non-voice section is calculated from the given voice content ratio S and the target companding ratio E so that the expression (1) is satisfied as the average speed ratio Vm1 = Es (fixed value) of the voice section. . However, Vm2 ≦ Es is a calculation condition.

  Thus, if the calculation pattern of the average speed ratio is different, the combination of the average speed ratio of the voice section and the average speed ratio of the non-voice section is different even if the voice content rate and the target companding ratio are the same. Therefore, which calculation pattern is to be selected is determined using the speed ratio calculation distribution according to the voice content rate and the target companding ratio. Hereinafter, the speed ratio calculation distribution will be described.

  FIG. 5 shows an example of the speed ratio calculation distribution. In FIG. 5, the vertical axis represents the voice content rate, the horizontal axis represents the target companding ratio, and the distribution of the calculation patterns a to c is shown. Here, the speed ratio calculation distribution is a value that the average speed ratio of the voice and non-voice sections can take while selecting a predetermined calculation pattern from the above-described calculation patterns and satisfying the above-described conditions for the selected calculation pattern. Can be obtained by setting In the speed ratio calculation distribution shown in FIG. 5, calculation patterns a to c are selected from the calculation patterns described above. In the calculation pattern a, the average speed ratio of the non-voice section is set to An = 4, and the average speed ratio Vm1 of the voice section can be set to 1.3 ≦ Vm1 ≦ 2. This possible value satisfies the calculation condition (Vm1 ≦ An) of the calculation pattern a. In the calculation pattern b, the average speed ratio of the voice section is set to Bs = 1.3, and the possible value of the average speed ratio Vm2 of the non-voice section is set to 1.3 ≦ Vm2 ≦ 4. This possible value satisfies the calculation condition (Vm2 ≧ Bs) of the calculation pattern b. In the calculation pattern c, the average speed ratio between the voice and non-voice sections is set as Vm1 = Vm2 (1 ≦ Vm1 ≦ 1.3). By selecting a calculation pattern in this way and setting possible values for the average speed ratio between voice and non-voice sections, the speed ratio calculation distribution of FIG. 5 can be obtained. When the voice content ratio and the target companding ratio are within the area of the calculation pattern a, the average speed ratio between the voice and the non-voice section is calculated by the calculation pattern a. The same applies to the calculation patterns b and c. In this way, which calculation pattern is calculated in accordance with the voice content rate and the target companding ratio is determined by the speed ratio calculation distribution. In the non-processable area on the left side of FIG. 5, the target companding ratio is extremely small with respect to the voice content rate, and the average speed ratio of the voice and non-voice sections is maximized within a range that can be heard by the user. However, it is an area where the target companding ratio cannot be achieved.

  In the speed ratio calculation distribution of FIG. 5, the higher the voice content rate, the higher the ratio calculated by the calculation pattern a with respect to the target companding ratio. In the calculation pattern a of FIG. 5, the average speed ratio of the non-voice section is set to the maximum value (An = 4). Thereby, in achieving the target companding ratio, the average speed ratio Vm1 of the voice section can be made as slow as possible.

  In the speed ratio calculation distribution of FIG. 5, the lower the voice content rate, the higher the ratio calculated by the calculation pattern b with respect to the target companding ratio. In the calculation pattern b of FIG. 5, the average speed ratio of the voice section is set to Bs = 1.3 (fixed value). Thereby, in achieving the target companding ratio, the average speed ratio Vm2 in the non-voice section can be made as slow as possible. As described above, in the speed ratio calculation distribution of FIG. 5, the correspondence between the voice content rate and the calculation method differs depending on the size of the voice content rate.

  When the target companding ratio is large, the calculation pattern c is selected. That is, the average speed ratio of voice and non-voice is made equal. When the target companding ratio is large, it is possible to reproduce more naturally when the voice and non-voice have the same average speed ratio. As described above, the speed ratio calculation distribution in FIG. 5 differs in the target companding ratio and the calculation method according to the size of the target companding ratio.

  Further, in the speed ratio calculation distribution shown in FIG. 5, possible values of the average speed ratio of the voice and non-voice sections are set so that the area is uniquely determined by the voice content ratio and the target companding ratio. That is, the possible values of the average speed ratio between the voice and non-voice sections are set so that the average speed ratio value is continuous at the boundary between adjacent calculation patterns. Specifically, in the calculation pattern a, the lower limit of the average speed ratio Vm1 in the voice section is 1.3, and the average speed ratio Bs in the voice section of the adjacent calculation pattern b is 1.3. As a result, the average speed ratio value continues at the boundary between the calculation patterns a and b. In the calculation pattern b, the average speed ratio Bs in the voice section is 1.3, and in the adjacent calculation pattern c, the maximum upper limit of the average speed ratio Vm1 in the voice section is 1.3. As a result, the average speed ratio value continues at the boundary between the calculation patterns b and c.

  In addition, it will be as follows if it demonstrates from a viewpoint of how the average speed ratio of an audio | voice and a non-voice area changes continuously as a target companding ratio becomes large. When the target companding ratio takes a value within the unprocessable region, the average speed ratio between the voice and non-voice sections is not calculated. When taking a value within the area of the calculation pattern a, the average speed ratio of the voice interval decreases from 2 to 1.3 as the target companding ratio increases. At this time, the average speed ratio in the non-speech section is constant at 4. When taking a value in the area of the calculation pattern b, the average speed ratio of the voice section is constant at 1.3, and the average speed ratio of the non-voice section decreases from 4 to 1.3 as the target companding ratio increases. . When taking a value in the area of the calculation pattern c, the average speed ratio of the voice and non-voice sections is the same value, but decreases from 1.3 to 1.

  In this way, the speed ratio calculation distribution is set so that the average speed ratio at the boundary where the calculation pattern changes becomes a continuous value, so that a rapid speed conversion occurs when the average speed ratio takes a discontinuous value. , It is possible to avoid the problem of uncomfortable feeling.

  FIG. 6 shows the target companding ratio, the average speed ratio of the voice section, and the average speed ratio of the non-voice section when the voice content rate is 0.5. As shown in FIG. 5 described above, the calculation pattern a is selected when the target companding ratio takes values from 0.375 to 0.501. When the target companding ratio takes values from 0.501 to 0.769, the calculation pattern b is selected. When the target companding ratio takes a value from 0.769 to 1, the calculation pattern c is selected. Here, as described above, values that can be taken by the average speed ratio are set in each calculation pattern shown in FIG. Therefore, the average speed ratio as shown in FIG. 6 is calculated by the above-described calculation pattern and equation (1).

  When the target companding ratio takes values of 0.1 and 0.3, as can be seen from the speed ratio calculation distribution shown in FIG. 5, it becomes an unprocessable region, so the speed ratio of the voice and non-voice sections is calculated. Not. When the target companding ratio takes values of 0.4 and 0.5, both are calculated by the calculation pattern a. In addition, when calculating with the calculation pattern a, it turns out that the average speed ratio of a voice area becomes small as the target companding ratio increases. When the target draw ratio takes values of 0.6 and 0.7, both are calculated by the calculation pattern b. In addition, when calculating by the calculation pattern b, it turns out that the average speed ratio of a non-voice area becomes small as the target companding ratio increases. When the target companding ratio takes values of 0.9 and 1.0, the average speed ratio of the voice and non-speech sections becomes equal, and the average speed ratio of the voice and non-speech sections decreases as the target companding ratio increases. You can see that

  Next, a method for calculating the termination speed ratio of the voice section will be described. The speed ratio condition setting unit 14 calculates the termination speed ratio of the voice section from the calculated average speed ratio of the voice section. FIG. 7 is a schematic diagram showing a change in speed ratio between voice and non-voice sections. The section length of the voice section 1 is shorter than the section length of the voice section 2. The vertical axis is the conversion speed ratio, and the horizontal axis is the elapsed time. The conversion speed ratio indicates a speed ratio used for the speed conversion processing of the speed conversion unit 17 and is indicated by a speed ratio in each section obtained by subdividing the voice and non-voice sections. A method for determining the conversion speed ratio will be described later. As shown in FIG. 7, even if the voice section length is different, the speed ratio at the end time of the voice section is the same. The speed ratio condition setting unit 14 calculates the speed ratio at the end time as the end speed ratio of the voice section. As is clear from FIG. 7, a speed ratio slower than the terminal speed ratio is set in the voice section. The terminal speed ratio Vend of the voice section is obtained by adding α to the average speed ratio Vm1 of the voice section. That is, the terminal speed ratio Vend = Vm1 + α. In addition, as a result of listening experiments, it was found that α is preferably 0.2 and Vend does not exceed 2.0. In FIG. 7, the speed ratio of the non-voice section is constant at the average speed ratio Vm2end.

  As described above, the speed ratio condition setting unit 14 receives the voice content rate and the target companding ratio, and calculates the average speed ratio of the voice section, the average speed ratio of the non-voice section, and the end speed ratio of the voice section. These are set as speed ratio conditions.

  The voice segment length calculation unit 15 receives the start and end times of the voice segment and calculates the voice segment length. The speed ratio determining unit 16 determines the conversion speed ratio between the voice and non-voice sections based on the speed ratio condition set by the speed ratio condition setting unit 14 and the voice section length. Here, the conversion speed ratio indicates the speed ratio used for the speed conversion processing of the speed conversion unit 17 as described above, and is indicated by the speed ratio in each section obtained by subdividing the voice and non-voice sections. . However, when the speed ratio in the speech section is made constant or when the speed ratio is switched at regular intervals, it is not always necessary to calculate the speech section length, so the speech section length calculation unit 15 does not have to be provided. Even if the voice section length calculation unit 15 is not provided, since the average speed ratio of the voice section is set according to the voice content rate, it is easier to hear than the conventional method. On the other hand, by providing the voice segment length calculation unit 15, the voice segment length is calculated, and the speed ratio is set for each segment of the voice segment, which makes it easier to hear.

  The speed conversion unit 17 receives the audio signal and performs speed conversion according to the conversion speed ratio determined by the speed ratio determination unit 16. As a method of speed conversion, for example, “Realization of high-quality voice speed conversion system by DSP” <Suzuki, Misaki, Society of Electronics, Information and Communication Engineers, Speech Research Group Material SP90-34, (1990.8.23)>, Japanese Patent No. 3189588 It is assumed that a known method described in the above is used. By such a method, reproduction at a slow speed ratio of 1 × speed or less and reproduction at a fast speed ratio of 1 × speed or more are possible. In addition, the speed conversion method is not limited to this, and a process that synthesizes sound, deletes or inserts a section, or performs a process that satisfies the conversion speed ratio determined by the speed ratio determination unit 16 is performed. Any method can be used. For example, assuming that the conversion speed ratio is 0.5, it is sufficient that the output playback time is doubled for a certain input section, and the sound is stretched, a silent section is added, or a new sound is synthesized. May be. As described above, the speed conversion unit 17 is associated with the relationship between the input and the output for a certain section, and is included as a speed conversion method as long as it performs processing that satisfies the conversion speed ratio.

  Hereinafter, with reference to FIG. 8, the process of the audio reproduction device according to the first embodiment will be described. FIG. 8 is a flowchart showing a flow of processing of the audio reproduction device according to the first embodiment.

  First, when a user gives an instruction to record content on an input device (not shown), an audio signal and a video signal of the content are stored in the storage unit 12. At this time, the voice / non-voice discriminating unit 11 discriminates a voice section and a non-voice section from the audio signal of the content (step S101). Note that the determination result determined in step S101 and the start and end times of the speech segment are also stored in the storage unit 12.

  After step S101, it is determined whether or not the user has instructed the user to play back the desired content (step S102). When there is an instruction from the user (Yes in step S102), the audio content rate calculation unit 13 calculates the audio content rate of the instructed content (step S103).

  After step S103, the user sets a target companding ratio with an input device (not shown) (step S104). The speed ratio condition setting unit 14 determines whether or not the target companding ratio set in step S104 is within an unprocessable region based on the voice content rate calculated in step S103 and the preset speed ratio calculation distribution. Is determined (step S105). If the target companding ratio is set in the unprocessable area (No in step S105), the process returns to step S104. In step S104, the speed ratio condition setting unit 14 resets the target companding ratio to a processable value. In the case of FIG. 5, the minimum target companding ratio that can be processed when the voice content is 0.5 is 0.375. Therefore, the speed ratio condition setting unit 14 resets the nearest region boundary value 0.375 as the target companding ratio. Note that the speed ratio condition setting unit 14 may not automatically reset the speed ratio condition setting unit 14 but may request the user to input the target companding ratio again.

  After step S105, the speed ratio condition setting unit 14 determines the average speed ratio of the voice interval and the non-voice interval based on the voice content rate calculated in step S103 and the target companding ratio set in steps S104 and S105. Are calculated, and the speed ratio condition is set (step S106). The method for calculating the speed ratio condition is as described above.

  After step S106, the speech segment length calculator 15 receives the start / end time of the speech segment and calculates the speech segment length (step S107). As shown in FIGS. 9 and 10, the length of the voice section includes various lengths within the same content, but greatly varies depending on the genre. FIG. 9 is a diagram showing the length of a voice segment included in a news program and its frequency. FIG. 10 is a diagram showing the length of a voice segment included in a baseball program and its frequency. 9 and 10, the horizontal axis represents the voice section length, and the vertical axis represents the frequency of occurrence during the program.

  Here, in the above-described prior art, the speed ratio is set only by the elapsed time from the start without considering the voice section length, and the speed ratio is gradually increased with the elapsed time when the voice section length is long, There was a problem that it was difficult to hear. On the other hand, in the present embodiment, by considering the voice section length, as shown in FIG. 7, the speed ratio is set so that the speed ratio at the end of the voice section becomes equal regardless of the length of the voice section length. Can be determined. As a result, the speed ratio of the voice section is gradually increased from the beginning, but the speed is increased only to the speed ratio at which the end can be heard.

  After step S107, the speed ratio determination unit 16 refers to the start / end time of the audio section stored in the storage unit 12, and determines whether or not it is an audio section every predetermined unit time in order from the start end of the content. Judgment is made (step S108). When it is determined that the voice segment is determined, the speed ratio determination unit 16 calculates the elapsed rate in the voice segment based on the start / end time of the voice segment and the voice segment length calculated in step S107 (step S109). The elapsed rate of the speech segment is obtained by dividing the elapsed time from the start end by the speech segment length, where 0 is the start end and 1 is the end of the speech segment.

平均圧伸比ｖｍ１は、音声区間の平均速度比Ｖｍ１の逆数である。終端圧伸比ｖｅは、終端速度比Ｖｅｎｄの逆数である。ここで、圧伸比変化量は、音声区間の平均圧伸比ｖｍ１を０と想定したとき、ｖｍ１に対して増減する量（図１１の網掛け部分の面積）を意味する。したがって、この量の和が０となるためには、図１１に示したように、ｘ＝０．５のときに変換圧伸比ｖｘが平均圧伸比ｖｍ１となるようにすればよい。平均圧伸比ｖｍ１は、音声区間の平均速度比Ｖｍ１から求まる値であり、終端圧伸比ｖｅは、終端速度比Ｖｅｎｄから求まる値である。したがって、式（３）を満たすように始端圧伸比ｖｓを設定すればよいこととなる。

なお、経過割合がｘのときの変換圧伸比ｖｘ、始端圧伸比ｖｓ、終端圧伸比ｖｅ、平均圧伸比ｖｍ１を速度比で表すと、式（４）のようになる。ここで、Ｖｘは経過割合がｘのときの変換速度比、Ｖｓは始端速度比、Ｖｅｎｄは上述した終端速度比、Ｖｍ１は上述した平均速度比を示す。

そして、式（２）に式（３）および（４）を代入すると、式（５）が得られる。

なお、速度変換後の音声区間長は、平均速度比Ｖｍ１で一様に変換した時間長と等しくなることから、式（６）が成り立つ。

このようにステップＳ１１０において、速度比決定部１６は、式（５）に音声区間の経過割合ｘを代入することで、音声区間の変換速度比Ｖｘを決定することができる。このステップＳ１１０で算出した音声区間の変換速度比は、上述した図７のような変化となる。つまり、音声区間の冒頭部分を遅くし、終端に向かって徐々に速めていくように、変換速度比を音声区間長に応じて変化させることができる。 After step S109, the speed ratio determination unit 16 determines the conversion speed ratio of the voice section from the elapsed rate of the voice section (step S110). Hereinafter, a specific processing example of step S110 will be described. As an example of the conversion speed ratio calculation process, there is a method of calculating the conversion speed ratio so that the sum of the amount of change in the drawing ratio with respect to the average drawing ratio in the voice section becomes zero. FIG. 11 is a diagram showing a change in the companding ratio in the voice section. In FIG. 11, x is the elapsed ratio, vx is the conversion companding ratio when the elapsed ratio is x, vs is the start end companding ratio, ve is the end companding ratio, and vm1 is the average companding ratio. Here, the change curve of the drawing ratio connecting the starting end drawing ratio vs and the ending drawing ratio ve is expressed by Expression (2).

The average companding ratio vm1 is the reciprocal of the average speed ratio Vm1 of the voice section. The terminal companding ratio ve is the reciprocal of the terminal speed ratio Vend. Here, when the average companding ratio vm1 of the voice section is assumed to be 0, the companding ratio change amount means an amount that increases / decreases with respect to vm1 (the area of the shaded portion in FIG. 11). Therefore, in order for the sum of the amounts to be 0, as shown in FIG. 11, the converted companding ratio vx should be the average companding ratio vm1 when x = 0.5. The average companding ratio vm1 is a value obtained from the average speed ratio Vm1 of the voice section, and the terminal companding ratio ve is a value obtained from the terminal speed ratio Vend. Therefore, the starting end companding ratio vs may be set so as to satisfy the expression (3).

In addition, when the elapsed ratio is x, the converted companding ratio vx, the starting end companding ratio vs, the end companding ratio ve, and the average companding ratio vm1 are expressed by a speed ratio as shown in Expression (4). Here, Vx is the conversion speed ratio when the elapsed ratio is x, Vs is the start speed ratio, Vend is the end speed ratio described above, and Vm1 is the average speed ratio described above.

Then, when Expressions (3) and (4) are substituted into Expression (2), Expression (5) is obtained.

Note that the voice section length after the speed conversion is equal to the time length uniformly converted with the average speed ratio Vm1, and therefore Expression (6) is established.

As described above, in step S110, the speed ratio determination unit 16 can determine the conversion speed ratio Vx of the voice section by substituting the elapsed ratio x of the voice section into the equation (5). The conversion speed ratio of the speech section calculated in step S110 changes as shown in FIG. That is, the conversion speed ratio can be changed in accordance with the length of the speech segment so that the beginning of the speech segment is delayed and gradually accelerated toward the end.

  If it is determined in step S108 that it is a non-speech segment, the speed ratio determination unit 16 determines the average speed ratio of the non-speech segment set by the speed ratio condition setting unit 14 as the conversion rate ratio from the start to the end of the non-speech segment. To do. That is, as shown in FIG. 7, the speed ratio determination unit 16 determines the conversion speed ratio from the start end to the end of the non-voice section so that the average speed ratio is constant.

  Next to steps S110 and S111, the speed ratio determination unit 16 determines whether or not the conversion speed ratio has been calculated up to the end of the content (step S112). If not, the process returns to step S108. In this way, the processing from step S108 to step S112 is repeated in the speed ratio determination unit 16 until the conversion speed ratio up to the end of the content is calculated. When it is determined in step S112 that the conversion speed ratio has been calculated up to the end of the content, the speed conversion unit 17 converts the speed of the audio signal according to the conversion speed ratio, and starts playing the audio signal after the speed conversion ( Step S113). The input device (not shown) receives an instruction as to whether or not to end the processing of this device (step S114). When the user performs the speed conversion process for other contents (No in step S114), the process returns to step S102.

  As described above, according to the audio reproduction device according to the present embodiment, the speed ratio condition corresponding to the content can be set by calculating the audio content rate of the content. Thereby, while achieving the target companding ratio, that is, the target time, the speed ratio between the voice section and the non-voice section can be set to the optimum speed ratio according to the content. As a result, whatever content audio signal is input, it can be played back at a speed that is easy to hear, and playback with reduced discontinuity of playback content and discomfort due to lack of information can be performed. it can.

  Further, according to the sound reproducing device of the present embodiment, the linear function is used as the function indicating the change in the drawing / compression ratio as shown in FIG. That is, the speed ratio is set with a linear line for the audio signal input to the apparatus. Here, Japanese is said to be a language of mora rhythm, and there is a tendency to speak so that each mora has the same length. Mora is a unit of length for speaking words, and in Japanese, it corresponds to beats counted in haiku and tanka. Speaking of “kana”, it corresponds to one character. Although it is desirable to change the speed ratio for each mora, since the speed ratio is calculated with a linear line for the input audio signal, it is sufficiently natural even without changing the speed ratio for each mora. Playback can be realized. Furthermore, the speed ratio from the beginning to the end of the speech section is finely switched by a linear function. As a result, the speed is changed at intervals shorter than the perceived time, and natural reproduction with less discomfort can be provided.

  In addition, according to the audio reproducing device according to the present embodiment, the speed ratio is set according to the audio section length from the start end to the end of the audio section. Accordingly, it is possible to prevent the speed ratio from becoming too fast near the termination time of the voice section and becoming difficult to hear without becoming a speed ratio faster than a preset termination speed ratio at the termination time of the voice section.

  In the above description, as shown in FIG. 11, the change in the drawing ratio is expressed by a linear function, but may be expressed by another function. For example, it may be an exponential function convex upward or convex downward. Also, for example, when the speed ratio that can be used in advance is limited, even if conversion is performed with two or several speed ratios, it is not possible to perform speed conversion according to the elapsed rate of the voice interval length. Reproduction with reduced naturalness can be provided. FIG. 12 is a diagram illustrating a case where a two-stage conversion speed ratio is calculated. In FIG. 12, it is more preferable that the first conversion speed ratio occupies a range of 20 to 30% from the start end with respect to the entire speech section. As a result, it became clear through listening experiments that more natural reproduction was realized. Further, for example, the speed ratio of the voice section may be a constant speed ratio as in the non-voice section. Even in this case, since an appropriate speed ratio is set depending on the audio content rate, it is not necessary to delete non-speech sections or perform extreme speedup as in the prior art, and provide easy-to-listen playback. Can do.

  In the above description, the speed ratio condition setting unit 14 uses the speed ratio calculation distribution shown in FIG. 5, but is not limited thereto. The user himself / herself may select a desired calculation pattern, set a possible value of the average speed ratio between the voice and non-voice sections to a desired value, and create a speed ratio calculation distribution. That is, the speed ratio calculation distribution used by the speed ratio condition setting unit 14 is not limited to that set in advance, and may be set by the user. For example, in FIG. 5, the average speed ratio of the voice section can be up to 2.0. However, elderly and language learners may wish to listen at an average speed ratio slower than 2.0. At this time, by setting a value that can be taken by the average speed ratio of the voice and non-voice sections so as not to exceed the average speed ratio desired by the user, it is possible to perform a reproduction process according to the user's listening ability. In addition, there are cases where elderly people and language learners want to listen at a slower playback speed than normal playback speed. The voice interval should be slower than the normal average speed ratio of 1.0, and the non-voice interval should be faster than the normal average speed ratio of 1.0 to fit within the normal playback time, or in a shorter time The speed ratio calculation distribution of the speed ratio condition setting unit 14 can be switched according to the purpose in order to answer the desire to view.

  Further, the speed ratio calculation distribution shown in FIG. 5 may be prepared in advance for each genre. In this case, which speed ratio calculation distribution is used is selected according to genre information such as EPG or a user instruction. Here, in addition to the audio content rate, the intensity of movement of the image varies depending on the genre. For example, in a genre with many still images such as documentaries, even if a non-voice section is played back at a high speed, information loss due to an increase in the speed of the image is small. Further, since the non-voice section can be played back at high speed, the voice section can be made close to 1 × speed. As a result, it is possible to perform reproduction that makes it easy to understand the contents of the program. Here, FIG. 13 shows an example of a speed ratio calculation distribution for a genre with many still images such as documentaries. As shown in FIG. 13, the speed ratio calculation distribution is composed of areas of calculation patterns a and b. Among these, in the calculation pattern a, the average speed ratio of the non-voice section is set to An = 4, and the possible value of the average speed ratio Vm1 of the voice section is set to 1 ≦ Vm1 ≦ 2. In the calculation pattern b, the average speed ratio of the voice section is set to Bs = 1, and the possible value of the average speed ratio Vm2 of the non-voice section is set to 1 ≦ Vm2 ≦ 4.

  FIG. 14 shows the target companding ratio, the average speed ratio of the voice section, and the average speed ratio of the non-voice section when the voice content rate is 0.5 in the speed ratio calculation distribution shown in FIG. When the target companding ratio takes values of 0.1 and 0.3, it becomes an unprocessable area as can be seen from the speed ratio calculation distribution shown in FIG. 13, so the speed ratio of the voice and non-voice sections is calculated. Not. When the target companding ratio takes values of 0.4, 0.5, and 0.6, both are calculated by the calculation pattern a. When the target companding ratio takes values of 0.7, 0.9, and 1, both are calculated by the calculation pattern b. As can be seen from FIG. 14, when the speed ratio calculation distribution shown in FIG. 13 is used, the difference between the average speed ratios of voice and non-voice increases because the calculation pattern is composed of two patterns. In other words, it means that the non-speech section can be speeded up and the voice section can be set to 1 × speed, so that the program contents can be reproduced more easily.

  In addition, in a genre where there are many scenes of intense movement such as sports, it is better not to make a large difference in the average speed ratio between voice and non-voice. This is because a genre with many scenes with a lot of movement has a greater influence on the content understanding of the program than a genre with little movement, so it improves the ease of hearing and visibility of non-voice segments. It is necessary. Here, FIG. 15 shows an example of the speed ratio calculation distribution for a genre with many scenes with a lot of motion such as sports. As illustrated in FIG. 15, the speed ratio calculation distribution includes regions of two calculation patterns a, two calculation patterns b, and calculation patterns c. Among these, in the leftmost calculation pattern a, the average speed ratio of the non-voice section is set to An = 3, and the possible value of the average speed ratio Vm1 of the voice section is set to 1.8 ≦ Vm1 ≦ 2.5. . In the calculation pattern b adjacent to the calculation pattern a, the average speed ratio of the voice section is set to Bs = 1.8, and the possible value of the average speed ratio Vm2 of the non-voice section is set to 2.5 ≦ Vm2 ≦ 3. . In the calculation pattern a adjacent to the calculation pattern b, the average speed ratio of the non-voice section is set to An = 2.5, and the possible value of the average speed ratio Vm1 of the voice section is set to 1.5 ≦ Vm1 ≦ 1.8. ing. In the calculation pattern b adjacent to the calculation pattern a, the average speed ratio of the voice section is set to Bs = 1.5, and the possible value of the average speed ratio Vm2 of the non-voice section is set to 1.5 ≦ Vm2 ≦ 2.5. ing. In the calculation pattern c adjacent to the calculation pattern b, the average speed ratio between the voice and non-voice sections is set as Vm1 = Vm2 (1 ≦ Vm1 ≦ 1.5).

  FIG. 16 shows the target companding ratio, the average speed ratio of the voice section, and the average speed ratio of the non-voice section when the voice content rate is 0.5 in the speed ratio calculation distribution shown in FIG. When the target companding ratio takes a value of 0.1 and 0.3, it becomes an unprocessable area as can be seen from the speed ratio calculation distribution shown in FIG. Not. When the target companding ratio takes values of 0.4 and 0.5, both are calculated by the calculation pattern a. When the target companding ratio takes a value of 0.6, it is calculated by the calculation pattern b. When the target companding ratio takes values of 0.7, 0.9, and 1, it is calculated by the calculation pattern c. As can be seen from FIG. 15, many calculation patterns are switched with respect to the target companding ratio. Thus, when the speed ratio calculation distribution shown in FIG. 15 is used, there is no significant difference between the average speed ratio of voice and non-voice. As a result, the ease of hearing and visibility of the non-voice section is improved. Further, the speed ratio of the non-voice section is set slightly later than the speed ratio calculation distribution shown in FIG. Thereby, it is possible to further improve the easiness to hear and view in the non-speech section for a genre with many scenes that are generated in the non-speech section. Thus, by preparing the speed ratio calculation distribution for each genre, more accurate speed conversion processing can be performed.

  In addition, not only for each genre, but also image information indicating the intensity of movement and a speed ratio calculation distribution according to acoustic characteristics may be prepared in advance. Such a speed ratio calculation distribution is effective, for example, when it is desired to individually control the speed of sound that the user wants to focus on, such as music or the sound of a specific person.

  In the above description, the calculation patterns a to c are used as the regions constituting the speed ratio calculation distribution. However, the calculation patterns d and e may be used according to the purpose. When music programs are to be played with an emphasis on music, the music is a non-speech segment, so the speed ratio of the non-speech segment should be made as slow as possible. Instead, it is necessary to increase the speed ratio of the voice section. Therefore, it is preferable to use the calculation pattern d described above for such a genre such as a music program. This is effective not only for music but also for reproduction of sound that the user wants to focus on. Further, when the user searches for a non-speech segment included in the content, it is desired to reproduce the speech segment as fast as possible. Therefore, in this case, it is preferable to use the calculation pattern e described above.

  Further, the speed ratio condition setting unit 14 obtains a statistical distribution as shown in FIG. 9 and FIG. 10 from the speech segment length obtained by the speech segment length calculation unit 15, and the speed based on the statistical distribution. A ratio calculation distribution may be used. The voice section length and the frequency of occurrence thereof indicate content attributes. For this reason, by using a speed ratio calculation distribution based on a statistical distribution, it is possible to perform a speed conversion process according to content attributes. For example, even if the audio content rate is the same, there are content with a short audio section length and a high occurrence frequency of the audio section, and content with a long audio section length and a high occurrence frequency of the audio section. In the latter content, it is assumed that the amount of information included per voice segment is large, and the burden on the user for understanding is high. Accordingly, for such content, a speed ratio calculation distribution that distributes a slower speed ratio more preferentially to the audio section is used. As described above, the speed ratio condition setting unit 14 may set the speed ratio condition using the speed ratio calculation distribution based on the statistical distribution of the voice section length. This is particularly effective for private content. Unlike broadcast contents, many private contents are not subjected to processing such as editing. Therefore, the content of audio and the length of an audio section vary greatly from content to content. Therefore, by preparing various speed ratio calculation distributions, it is possible to set an appropriate speed ratio condition even when the content of the voice section length or the voice content rate varies greatly between contents such as private contents.

なお、Ｓは音声含有率、Ｖｍ１は音声区間の平均速度比、Ｖｍ２は非音声区間の平均速度比、Ｅは目標圧伸比を示す。ｎ番目の音声区間の音声区間長をＭ（ｎ）とし、ｎ番目の音声区間に後続する非音声区間の非音声区間長をＮ（ｎ）とする。音声区間長と同じ長さの非音声区間長を設けるため、音声区間の平均比Ｖｍ１は式（７）のように表せる。また発音練習を行うには、音声区間と同じ長さの非音声区間を必要とする。このため、後続の非音声区間の速度比は、音声区間長に応じて算出する必要があるため、式（８）のようになる。このように、音声区間長と同じ長さの非音声区間を設け、一定の時刻内で再生するという発音練習モードにおいても、音声含有率や音声区間長を利用することで、適切な速度比を設定することが可能になる。なお、今回は音声区間と同じ長さの非音声区間を設けたが、非音声区間の長さは学習の進み具合などに応じて変更してもよい。このように、語学学習用に新たにコンテンツを作成しなくても、音声含有率と音声区間長の利用によって、発音練習に適した速度で音声を提示することが可能になる。また、学習に費やしたい時間を始めに指定することで、コンテンツ長から圧伸比を算出し、学習時間内におさまるように速度を制御することが可能になる。また、学習のレベルに応じて速度比を変えることも可能となる。 In the above description, only the case where the target companding ratio is 0 to 1 has been described. However, the playback time after speed control is the same as or longer than the normal playback time. Similarly, when the target companding ratio is 1 or more such as viewing and reproducing, the speed ratio condition can be set by preparing a speed ratio calculation distribution in advance. In addition, it is possible to perform output control in the pronunciation practice mode in which a non-speech segment having the same length as the speech segment length is provided for each voice segment to encourage pronunciation practice. For example, in the pronunciation practice mode, assuming that the average speed ratio of the non-speech section is determined by the length of the previous speech section, the speed ratio of the speech section and the non-speech section can be calculated from the following equation.

S represents the voice content rate, Vm1 represents the average speed ratio of the voice section, Vm2 represents the average speed ratio of the non-voice section, and E represents the target companding ratio. Let M (n) be the voice section length of the nth voice section, and N (n) be the non-voice section length of the non-voice section that follows the nth voice section. Since a non-speech segment length having the same length as the speech segment length is provided, the average ratio Vm1 of the speech segment can be expressed as in Expression (7). In order to practice pronunciation, a non-speech segment having the same length as the speech segment is required. For this reason, since it is necessary to calculate the speed ratio of the subsequent non-speech segment according to the speech segment length, Equation (8) is obtained. In this way, even in the pronunciation practice mode where a non-speech segment with the same length as the speech segment length is provided and played within a certain time, an appropriate speed ratio can be obtained by using the speech content rate and the speech segment length. It becomes possible to set. Although a non-speech segment having the same length as the speech segment is provided this time, the length of the non-speech segment may be changed according to the progress of learning. As described above, even when content is not newly created for language learning, it is possible to present the voice at a speed suitable for pronunciation practice by using the voice content rate and the voice section length. In addition, by designating the time to be spent for learning first, the companding ratio is calculated from the content length, and the speed can be controlled so as to fall within the learning time. It is also possible to change the speed ratio according to the learning level.

なお、Ｓは音声含有率、Ｖｍ１は音声区間の平均速度比、Ｖｍ２は非音声区間の平均速度比、Ｅは目標圧伸比を示す。コンテンツに含まれる音声区間の総数をＵとし、コンテンツの全長をＱとしている。Ｐは音声区間を再生後に、書き込み行の変更など書き起こし作業用に必要な時間として設けた一定時間である。従って作業によってはＰの値は０でも構わない。非音声区間は再生されないため、非音声区間の速度比Ｖｍ２は式（１０）のようになる。ここで、音声区間の平均速度比Ｖｍ１をユーザの能力に応じた書き込み速度に設定すれば、書き込み作業中に一時停止や早送り、巻き戻しなどの機器操作をすることなく、書き起こし作業が可能となる。このように速度条件設定部１４で音声区間と非音声区間の平均速度比を設定することで、ユーザの書き込み能力に応じた速度比で音声区間のみを再生するような書き起こし作業モードが可能となる。 As another purpose of use, a transcription mode used when transcription of speech can be considered. This case can also be handled by changing the speed ratio calculation distribution of the speed ratio condition setting unit 14 in the same manner. To transcribe the voice, the writing ability of the writer, for example, how many characters can be written in a certain time when filling in paper, how many types can be written in a certain time when typing on the keyboard It is necessary to change the playback speed according to the ability. If the playback speed is faster than the writing capability, the writing part is immediately overtaken and the part before the writing part is reproduced. Therefore, operations such as pause and rewind are required. In addition, since such a reproduction control operation interrupts the writing process, it wastes time twice. If the playback speed is slower than the writing ability, it will not be overtaken by the playback part, but after writing, there will be a waiting time until the next sound starts and there will be no change in wasted time. Therefore, when transcription is performed, it is necessary to reproduce at a speed according to the user's writing ability. In the conventional method, since the length of the voice segment and the voice content rate are unknown, the non-speech segment is played back at the same speed as the voice segment, or it has not been known in advance how long it takes when the speed is slowed down. If the following processing is performed by the speed condition setting unit 14 this time, it is possible to reproduce the voice section so that the voice section is easy to hear and the non-voice section is omitted. The average speed ratio of the voice section can be expressed as in equation (9).

S represents the voice content rate, Vm1 represents the average speed ratio of the voice section, Vm2 represents the average speed ratio of the non-voice section, and E represents the target companding ratio. The total number of audio sections included in the content is U, and the total length of the content is Q. P is a fixed time set as a time required for a transcription work such as a change of a writing line after reproduction of a voice section. Therefore, the value of P may be 0 depending on the work. Since the non-speech section is not reproduced, the speed ratio Vm2 of the non-speech section is expressed by Equation (10). Here, if the average speed ratio Vm1 of the voice section is set to a writing speed according to the user's ability, the transcription work can be performed without performing a device operation such as pause, fast forward, and rewind during the writing work. Become. In this way, by setting the average speed ratio between the voice section and the non-voice section by the speed condition setting unit 14, a transcription work mode in which only the voice section is reproduced at a speed ratio according to the user's writing ability is possible. Become.

  In addition, at the start of learning, the voice interval is played back at a slow speed, and the total of the learning time becomes a predetermined time while gradually increasing the speed ratio according to the elapsed learning time. Control for listening practice that controls the speed is also possible. For example, in one learning, it is also possible to perform control such that the speech section is reproduced at a slower speed at the start of learning and gradually the playback speed of the speech section is increased at the start of learning. Further, during the long-term learning, the playback speed of the voice section in the current learning may be controlled based on the elapsed time from the time when the learning is first started or the previous operation history. In this way, the playback speed of audio and non-audio segments can be changed, pauses, silence and audio can be added immediately before and after the audio segment, and the time of the entire content can be adjusted while inserting a screen. It is possible to cope with this.

(Second Embodiment)
With reference to FIG. 17, an audio reproducing apparatus according to the second embodiment of the present invention will be described. FIG. 17 is a block diagram illustrating a configuration example of an audio reproduction device according to the second embodiment. In FIG. 17, the audio reproduction device includes an audio non-audio discrimination unit 11, an audio content rate prediction unit 18, a speed ratio condition setting unit 14, an audio section length prediction unit 19, a speed ratio determination unit 16, a speed conversion unit 17, and The drawing ratio calculating unit 20 is configured. The present embodiment is different from the audio reproduction device according to the first embodiment in that reproduction processing in which the reproduction speed of the audio signal is changed is performed in real time. Hereinafter, different points will be mainly described.

式（１１）において、初期値Ｘ（０）＝Ｙ（０）は、一般的なコンテンツの音声含有率の平均値とする。また音声含有率Ｘ（ｚ）が算出されるまでの間予測音声含有率Ｙ（ｚ）はＹ（０）を維持するとする。図１８に、式（１１）で示される予測音声含有率Ｙ（ｚ）を求める方法を模式的に示す。図１８に示すように、現時点より過去数分の音声含有率は、フレーム毎に移動しながら算出される。また、予測音声含有率Ｙ（ｚ）やＹ（ｚ−１）がセクション毎に予測される。図１８に示すように、音声含有率予測部１８は、一つ前のセクションでの予測音声含有率Ｙ（ｚ−１）と、現時点より過去数分の音声含有率Ｘ（ｚ−１）を用いて予測している。 The voice content rate prediction unit 18 calculates the voice content rate for the past several times from the present time based on the discrimination result for each frame output from the voice non-speech discrimination unit 11. And the audio | voice content rate in the section advanced one from the present time is estimated using the calculated audio | voice content rate. If the voice content rate for the past several minutes from the present time is X (z−1), the predicted value Y (z) of the voice content rate is expressed by Expression (11). Y (z-1) is the predicted value of the speech content rate in the previous section. Hereinafter, the predicted value of the voice content rate is referred to as a predicted voice content rate. α is a value from 0 to 1, and an optimum value is obtained by simulation.

In equation (11), the initial value X (0) = Y (0) is the average value of the audio content rate of general content. Further, it is assumed that the predicted speech content rate Y (z) is maintained at Y (0) until the speech content rate X (z) is calculated. FIG. 18 schematically shows a method for obtaining the predicted speech content Y (z) represented by the equation (11). As shown in FIG. 18, the audio content rate for the past several minutes from the present time is calculated while moving from frame to frame. Further, the predicted voice content rate Y (z) and Y (z-1) are predicted for each section. As shown in FIG. 18, the speech content rate prediction unit 18 calculates the predicted speech content rate Y (z−1) in the previous section and the speech content rate X (z−1) for the past several minutes from the present time. Predict using.

  Here, if the calculation time of the voice content rate X (z) is too short, the voice content rate becomes 1 if the voice section length is long. Further, only a short pause section between voices may be extracted, and an extreme value may be taken, such as the voice content rate being close to zero. On the other hand, if the calculation time is too long, it is smoothed and cannot be used for predicting the voice content rate. Therefore, the calculation time of the audio content rate X (z) needs to appropriately represent the degree of collection of the audio sections of the content, and as a result of the experiment, it was found that 1 minute or more is desirable. Therefore, in the above description, the past number is set. FIG. 19 shows an example of calculation results of the voice content rate X (z) and the predicted voice content rate Y (z). In this graph, the vertical axis represents the audio content rate and the horizontal axis represents the elapsed time of the program. FIG. 19 shows the audio content rate X (z) and the predicted audio content rate Y (z) calculated by equation (11) for a 30-minute news program. As shown in FIG. 19, it can be seen that the predicted voice content rate Y (z) changes to substantially the same value as the actual voice content rate X (z).

The drawing ratio calculation unit 20 calculates the drawing ratio in section units. Specifically, first, the companding ratio at the current time t is calculated. The companding ratio at the current time t is obtained by dividing the output data amount output from the speed conversion unit 17 at the current time t by the input data amount input to the speed conversion unit 17. Next, the companding ratio calculating unit 20 determines whether or not the current time t has reached the section boundary. When the section boundary is reached, in order to set the speed ratio condition in the next section, the companding ratio in the next section (hereinafter referred to as section companding ratio) is calculated. The section drawing ratio defines how much the next section is to be converted. The companding ratio calculation unit 20 calculates the target companding ratio set in advance by the user or the device and the companding ratio at the current time t using Expression (12). In Expression (12), Rt is the target companding ratio, R (t) is the companding ratio at the current time t, the time length of the next section is T (z), and Rs (z) is the companding ratio of the next section. And

  The speed ratio condition setting unit 14 receives the predicted voice content calculated by the voice content prediction unit 18 and the section companding ratio calculated by the companding ratio calculating unit 20. The speed ratio condition setting unit 14 sets the speed ratio condition for each section in the same manner as the speed ratio condition setting unit 14 described in the first embodiment.

  Here, in the case of performing the above-described sequential processing, in the conventional technique, when a local voice section bias occurs, processing is performed so as to achieve the reproduction time at the location where the bias occurs, so the non-voice section The extreme deletion and speeding up of the system occurred locally. In contrast, in this embodiment, the section companding ratio is calculated in section units. The speed ratio condition setting unit 14 sets speed ratio conditions in section units. That is, the speed ratio condition is updated for each section by the section companding ratio. As a result, it is possible to carry forward to the next and subsequent sections without achieving the playback time at the location where the bias occurs, so that it is possible to perform sequential processing that achieves playback that is easy to hear while achieving the target time. it can.

Ｗ（ｚ）は、セクションＺにおける、調整後の予測音声含有率である。以下、調整音声含有率と呼ぶ。Ｗ（ｚ−１）は、セクションＺの１つ前のセクションにおける調整音声含有率である。Ｙ（ｚ）は、セクションＺにおける予測音声含有率である。Ｙ（ｚ−１）は、１つ前のセクションにおける予測音声含有率である。γは、加算する際の係数である。 Note that if the predicted speech content rate is used as it is, an increase or decrease in the predicted speech content rate and an increase or decrease in the average speed ratio are directly connected. An increase in the average speed ratio at a portion where the predicted speech content is high may affect hearing. This is because, for example, assuming that the utterance speed is the same, the higher the voice content rate, the greater the amount of information, making it difficult to understand the sentence. Therefore, the predicted speech content ratio may be adjusted as follows so that the average speed ratio is decreased when the predicted speech content ratio increases and the average speed ratio is increased when the predicted speech content ratio decreases.

W (z) is the adjusted predicted speech content rate in section Z. Hereinafter, it is referred to as the adjusted voice content rate. W (z-1) is the adjusted audio content rate in the section immediately before section Z. Y (z) is the predicted speech content in section Z. Y (z-1) is the predicted speech content in the previous section. γ is a coefficient for addition.

  By setting the speed ratio condition using this adjusted voice content rate, if the predicted voice content rate rises above the predicted voice content rate in the previous section, the average speed ratio of the voice interval is lowered. When the predicted voice content rate is the same as the predicted voice content rate in the previous section, the average speed ratio of the voice section does not change.

  By using such an adjusted audio content rate, it is possible to make it slower in a section with a high audio content rate, and it is possible to set a speed condition according to the amount of information. Note that, by making such an adjustment, there is a possibility that the section expansion ratio cannot be achieved and an error may occur, but the error can be eliminated in the following sections as shown in Expression (12). Thereby, the target companding ratio can be achieved. For example, as shown in FIG. 19, the predicted speech content rate is not constant and may be high or low. Therefore, even if the playback speed is slowed down in the section where the predicted speech content rate is high and the difference from the target companding ratio is widened, this difference can be eliminated in the sections where the predicted speech content rate is low thereafter.

  The speech segment length prediction unit 19 calculates the speech segment length from the past discrimination result of the speech non-speech discrimination unit 11 and predicts the speech segment length. The predicted value of the speech segment length is predicted not by using the predicted value of each speech segment length but by using the average speech segment length when speaking a sentence as a representative value. The length of each speech segment is difficult to predict due to various factors such as speaker change and conversation content. Therefore, speed ratio control of a voice section suitable for content is performed using an average voice section length as a predicted value. As shown in FIG. 9 and FIG. 10, the distribution of the voice section length is greatly different for each genre. FIG. 9 shows the frequency of each voice section length in a news program. The frequency gradually decreases to about 4000 ms with a peak at 500 ms. On the other hand, FIG. 10 shows the frequency of each voice section length in the baseball program, but although the peak is also 500 ms, it can be seen that the frequency decreases rapidly. Due to this difference in the way of reduction, the impression of viewing these programs seems to be that a longer audio segment continues in a news program and a shorter audio segment continues in a baseball program. For this reason, if speed ratio control of a voice section is performed with a fixed length without predicting the voice section length, the speed ratio of the voice section may be unnecessarily slow or too fast. This is because the speech non-speech discriminating unit 11 sequentially discriminates between speech and non-speech, so that the end time of the speech section cannot be grasped. Also, the difference in the voice section length is due to the difference in the speaker and the conversation content, and differs for each content. For this reason, it is necessary to predict the speech interval length.

なお、音声区間には、「はい」や「うん」など相槌や、「えー」などのフィラーなどが含まれる。これらは言語的な理解が容易なため、速度比に関わらず聞き取り易い。そこで、式（１４）の音声区間長の予測値Ｌ（ｎ）の算出には、所定の閾値以上の音声区間長をもつ音声区間を利用するとした。ここでは、所定の閾値の一例として、１０００ｍｓを採用する。 Therefore, the predicted value L (n) of the speech section length of the nth (n is a natural number) speech section from the start of processing is the prediction of the speech section length of the previous (n-1) th speech section. From the value L (n-1) and the actual measurement value M (n-1), it is expressed by the equation (14).

Note that the speech section includes a companion such as “Yes” or “Yes”, a filler such as “Eh”, and the like. These are easy to hear regardless of speed ratio because they are easy to understand linguistically. Therefore, it is assumed that a speech segment having a speech segment length equal to or greater than a predetermined threshold is used to calculate the speech segment length prediction value L (n) in Expression (14). Here, 1000 ms is adopted as an example of the predetermined threshold.

  The speech segment length predicted based on the equation (14) is shown in FIG. In FIG. 20, the vertical axis represents the voice section length, and the horizontal axis represents the elapsed time. In FIG. 20, the speech segment length is shown at the position of the start time of the speech segment length on the X axis. Further, in FIG. 20, an actual measurement value M (n) of the speech segment length, an actual measurement value M (n−1) of the immediately preceding speech segment length, and a predicted speech segment length L (n) are illustrated. Thus, by calculating the predicted speech segment length from the speech segment length excluding the conflicts and fillers, it is possible to set a speed ratio suitable for a speech segment having a long speech segment length that is difficult to hear at a high speed ratio. . In addition, since the voice section length is sequentially calculated according to the content, the speed ratio can be set according to the difference in the voice section length distribution. The speed ratio determination unit 16 determines the conversion speed ratio between the speech and non-speech sections based on the sequential speed ratio condition and the predicted speech section length.

  Hereinafter, with reference to FIG. 21, the processing of the audio reproduction device according to the second embodiment will be described. FIG. 21 is a flowchart showing a flow of processing of the audio reproduction device according to the second embodiment.

  First, it is determined whether or not an instruction to reproduce content by a user is received at an input device (not shown) (step S201). When the user gives an instruction to reproduce the content, an audio signal is input to the voice / non-voice discrimination unit 11. The voice / non-speech discrimination unit 11 discriminates a speech segment and a non-speech segment for each frame of the audio signal of the input content (step S202).

  After step S202, the speech content rate prediction unit 18 calculates the speech content rate for the past several times from the current time based on the discrimination result for each frame output from the speech non-speech discrimination unit 11, and calculates the calculated speech content rate. The speech content rate in the section advanced by 1 from the current time is predicted (step S203).

  Following step S203, the speed ratio condition setting unit 14 receives the predicted speech content rate calculated by the speech content rate prediction unit 18 and the section companding ratio calculated by the companding ratio calculation unit 20 as input. The speed ratio condition is set to (Step S204). Further, the speech segment length prediction unit 19 calculates a speech segment length from the past discrimination result of the speech non-speech discrimination unit 11, and predicts the speech segment length (step S205).

  Next to step S205, the speed ratio determination unit 16 refers to the start / end time of the speech segment output from the speech non-speech discrimination unit 11, and determines whether or not the speech segment is a predetermined unit time. (Step S206). If it is determined that the voice segment is determined, the speed ratio determination unit 16 calculates the elapsed rate in the voice segment (step S207). The elapsed rate of the speech segment is obtained by dividing the elapsed time from the start end by the speech segment length, where 0 is the start end and 1 is the end of the speech segment. In the present embodiment, since voice / non-voice discrimination is performed sequentially, the start time of the voice section can be grasped, but the end time of the voice section is not known at the present time. Therefore, the predicted speech segment length predicted by the speech segment length prediction unit 19 is used as the speech segment length. Thereby, the speed ratio determination unit 16 can calculate the elapsed ratio in the voice section. In addition, since the predicted voice section length is used as the voice section length instead of the actual value, it does not necessarily match the elapsed ratio of the actual voice section. Therefore, there is a possibility that the end time of the voice section may be reached even if the elapsed ratio of the voice section is 1 or less. Therefore, the speed ratio determination unit 16 determines whether or not the passage ratio of the voice section exceeds 1 (step S208). When the ratio does not exceed 1, the speed ratio determination unit 16 determines the conversion speed ratio of the voice section from the elapsed ratio of the voice section (step S209). Since the process of step S209 is the same as the process described in the speed ratio determination unit 16 of the first embodiment, the description thereof is omitted. When the elapsed ratio exceeds 1 in step S208, the process proceeds to step S210, and the terminal speed ratio, which is the speed ratio condition, is calculated as the conversion speed ratio. In this case, since the voice section length predicted by the voice section length prediction unit 19 is exceeded, it is necessary to calculate the termination speed ratio of the voice section as the conversion speed ratio.

  When it is determined in step S206 that it is a non-speech section, the speed ratio determination unit 16 determines the conversion speed ratio of the non-speech section (step S211). Next to steps S209, S210, and S211, the speed ratio determination unit 16 determines whether or not the conversion speed ratio has been determined up to the end time of the content to be converted (step S212). If it is not the end time, the process returns to step S202. In this way, the processing from step S202 to S212 is repeated for each section in the speed ratio determination unit 16 until the conversion speed ratio up to the end time of the content to be speed converted is calculated. When it is determined in step S212 that the conversion speed ratio has been calculated up to the end time of the content, the input device (not shown) receives an instruction as to whether or not to end the processing of this device (step S213). When the user performs the speed conversion process for other contents (No in step S213), the process returns to step S202.

  As described above, according to the audio reproduction device according to the present embodiment, speed conversion can be performed while performing processing in real time. Compared to the first embodiment, predicted values are used for the voice content rate and the voice section length. For this reason, an error from the actual measurement value occurs, but this error is eliminated by the section companding ratio set by the companding ratio setting unit 28. Thereby, according to the audio reproducing device according to the present embodiment, the speed conversion is performed without deleting the section or extremely speeding up the audio section while achieving the target companding ratio while performing the processing in real time. be able to.

(Third embodiment)
With reference to FIG. 22, an audio reproducing apparatus according to the third embodiment of the present invention will be described. FIG. 22 is a block diagram illustrating a configuration example of an audio reproduction device according to the third embodiment. In FIG. 22, the audio reproduction device includes an audio non-audio discrimination unit 11, an audio content rate prediction unit 18, a speed ratio condition setting unit 14, an audio section length prediction unit 19, a speed ratio determination unit 16, a speed conversion unit 17, a pressure An elongation ratio calculation unit 20 and a statistic calculation unit 21 are included. In the present embodiment, a statistic calculation unit 21 is newly added to the audio reproduction device according to the second embodiment, and the processing of the speed ratio condition setting unit 14 is different. Hereinafter, the processing of the statistic calculation unit 21 and the speed ratio condition setting unit 14 will be mainly described.

  The statistic calculator 21 calculates a statistic for correcting the upper limit speed ratio of the voice section. For example, the audio content rate from the beginning of content to the current time is used. Such voice content is hereinafter referred to as long-term voice content. FIG. 23 shows the time variation of the long-term audio content rate for each content. In FIG. 23, the vertical axis represents the long-term voice content rate, and the horizontal axis represents the elapsed time from the start (0 minutes). Moreover, the predicted audio | voice content rate for every content is shown in FIG. The predicted voice content rate is the same as the predicted voice content rate described in the second embodiment. Here, the calculation interval is 1 minute. The graph of the predicted speech content rate is a graph in which the valleys are clear, reflecting portions where speech sections are dense or sparse. The graph of the long-term audio content rate is substantially flat, although there is some variation near the beginning, and is a graph close to the audio content rate for the entire content used in the first embodiment.

  Therefore, it is considered that the upper limit speed ratio of the voice section is corrected using the long-term voice content rate. By sequentially correcting the upper limit speed ratio of the speech section, it is possible to prevent the speed ratio of the speech section from being excessively increased even when the predicted speech content rate is locally increased.

  The speed ratio condition setting unit 14 includes a predicted voice content rate calculated by the voice content rate prediction unit 18, a long-term voice content rate calculated by the statistic calculation unit 21, and a section pressure calculated by the companding ratio calculation unit 20. The stretch ratio is input. In step S204 of FIG. 21 described above, the speed ratio condition setting unit 14 performs the process shown in FIG. FIG. 25 is a flowchart showing processing of the speed ratio condition setting unit 14 according to the third embodiment.

  In FIG. 25, the speed ratio condition setting unit 14 determines whether or not the input voice content rate is the predicted voice content rate (step S301). When the predicted voice content rate is input, the process proceeds to step S302, and the speed ratio condition setting unit 14 calculates the average speed ratio of voice and non-speech using the predicted voice content rate. Further, the speed ratio condition setting unit 14 calculates a terminal speed ratio from the calculated average speed ratio of the voice section (step S303). Note that the end speed ratio based on the predicted voice content is Vend1. The processes in steps S302 and S303 are the same as those in the first embodiment described above.

  On the other hand, when the predicted speech content rate is not input in step S301, that is, when the long-term speech content rate is input, the speed ratio condition setting unit 14 calculates the average speed ratio of speech and non-speech using the long-term speech content rate. (Step S304). Further, the speed ratio condition setting unit 14 calculates a terminal speed ratio from the calculated average speed ratio of the voice section (step S305). Note that the end speed ratio based on the long-term voice content is Vend2. The processes in steps S304 and S305 are the same as those in the first embodiment described above.

  Next to step S305, the speed ratio condition setting unit 14 sets the termination speed ratio Vend2 of the voice section calculated based on the long-term voice content rate as the upper limit speed ratio (step S306). Next to steps S303 and S306, the speed ratio condition setting unit 14 compares the terminal speed ratio Vend1 based on the predicted voice content rate and the upper limit speed ratio Vend2 based on the long-term voice content rate (step S307). When the terminal speed ratio Vend1 exceeds the upper limit speed ratio Vend2 (Vend1> Vend2), the speed ratio condition setting unit 14 corrects the terminal speed ratio Vend1 to the upper limit speed ratio Vend2 (step S308). Along with this correction, the speed ratio condition setting unit 14 also corrects the average speed ratio of the speech section to a value calculated by the long-term speech content rate. That is, among the speed ratio conditions representing the three speed ratios, the average speed ratio of speech and non-speech sections and the end speed ratio of speech sections, the value calculated with the predicted speech content only for the average speed ratio of non-speech sections Use. For the average speed ratio and end speed ratio of the other speech sections, values obtained from the long-term speech content rate are used.

  Thus, even when the average speed ratio of the voice section can be set to be higher by performing the correction of the average speed ratio and the terminal speed ratio of the voice section sequentially, the predicted voice content rate can be increased locally, Playback at a speed ratio that is easy to hear can be performed.

  The speech segment length prediction unit 19 calculates the predicted speech segment length from the past discrimination result of the speech non-speech discrimination unit 11. In this embodiment, the maximum value of the speech segment length is used as the predicted speech segment length. This is because speed ratio control can be performed up to the end without leaking any speech section from the viewpoint of ease of listening.

式（１５）において、ｍａｘはコンテンツに含まれる音声区間長のうち最大の音声区間長を複数のコンテンツについて平均した値である。事前にジャンル情報が得られる場合は、ジャンル情報毎に上記平均値を算出し、テーブルを用意しておく。 As shown in FIG. 26, the distribution of the voice section length varies greatly depending on the elapsed time. For this reason, it is necessary to predict the speech section length. FIG. 26 is a diagram showing the distribution of the measured value of the speech section length, the measured value of the immediately preceding speech section length, and the predicted speech section length. In FIG. 26, the vertical axis indicates the audio section length, and the horizontal axis indicates the elapsed time from the beginning of the content. The voice section length is displayed at the start time of the voice section. Here, the measured value of the nth speech segment length is M (n). The predicted speech interval length Lm (n) is expressed as in Expression (15) to Expression (17).

In Expression (15), max is a value obtained by averaging the maximum audio section length among a plurality of contents among the audio section lengths included in the content. When genre information is obtained in advance, the average value is calculated for each genre information, and a table is prepared.

なお、βは指数関数でもよく、経過時間ｔの減少関数であればよい。 β is a value set so that the predicted speech segment length Lm decreases with the elapsed time until the next speech segment. If the start time of the (n-1) th speech section is 0 and the start time of the nth speech section is t, it can be expressed as in equation (18). k takes a positive value.

Note that β may be an exponential function or a decreasing function of the elapsed time t.

  The predicted speech section length predicted by the equations (15) to (18) is as shown in FIG. As shown in FIG. 26, there are many cases in which the predicted speech interval length is longer than the actual measurement value of the speech interval length. By using this value when calculating the speed ratio, there is an effect that the rate of conversion by the termination speed ratio is reduced at the termination time of the speech section, and the average speed ratio of the speech section is further lowered. As a result, it is possible to provide playback that is easy to hear.

(Fourth embodiment)
With reference to FIG. 27, an audio reproducing apparatus according to the fourth embodiment of the present invention will be described. FIG. 27 is a block diagram illustrating a configuration example of an audio reproduction device according to the fourth embodiment. In FIG. 27, the audio reproduction device includes an audio non-audio discrimination unit 11, a temporary storage unit 22, an audio content rate calculation unit 13, a speed ratio condition setting unit 14, an audio section length calculation unit 15, a speed ratio determination unit 16, a speed A conversion unit 17 and a companding ratio calculation unit 20 are included. The present embodiment is provided with a temporary accumulation unit 22 having a smaller accumulation amount than the accumulation unit 12 and a companding ratio calculation unit 20 described in the second embodiment with respect to the audio reproduction device according to the first embodiment. It is different. Hereinafter, different points will be mainly described.

  The temporary storage unit 22 includes a readable / writable recording medium such as a hard disk, a DVD, or a memory medium (for example, an SD card). The temporary storage unit 22 stores the same audio signal that is input to the voice / non-voice discrimination unit 11 for one section or several sections. Then, after the audio signal stored in the temporary storage unit 22 is subjected to speed conversion processing, the audio signal of the section subjected to the speed conversion processing is deleted, and the audio signal of a new section is stored. Here, the section according to the present embodiment may be a section delimited by a predetermined event as well as a section delimited by a predetermined interval. For example, if the event is a CM, there are two types of sections: a CM section and a program section sandwiched between CMs. If the event is music, there are two types of sections: a music section and a section between music and music. The section may be a section designated by the user.

  As in the first embodiment, when the audio signal and the video signal constituting one section are stored in the temporary storage unit 22, a determination process is performed in the voice / non-voice determination unit 11, and the determination result of the section And the start / end time of the voice section are also stored in the temporary storage unit 22. The temporary storage unit 22 stores the audio signal and the video signal, the determination result, and the start / end time of the audio section in association with each other. The format of the audio signal and the video signal may be any format. Moreover, the audio reproduction device according to the present embodiment may further include the above-described storage unit 12. In this case, the audio signal and voice / non-speech discrimination result stored in the storage unit 12 in units of contents are read out in units of sections and stored in the temporary storage unit 22.

  By providing such a temporary storage unit 22, it is possible to calculate the actual voice section length (actually measured value) in the stored section. This makes it possible to control the speed according to the actual voice interval. In addition, since the actual audio content rate of the accumulated section can be obtained, there is less local variation compared to the case where the predicted audio content rate described in the second embodiment is used, and the audio content rate of the entire content is reduced. And close to the value.

  The voice content rate calculation unit 13 calculates the voice content rate of the section from the determination result stored in the temporary storage unit 22 and the start / end time of the voice section. The sum of the voice section lengths included in this section is calculated and divided by the time length of the entire section (hereinafter referred to as section length) to obtain the voice content rate of this embodiment.

  Hereinafter, with reference to FIG. 28, the process of the audio reproduction device according to the fourth embodiment will be described. FIG. 28 is a flowchart showing a flow of processing of the audio reproduction device according to the fourth embodiment.

  First, in an input device (not shown), it is determined whether or not the user gives an instruction to reproduce desired content (step S401). When the user gives an instruction, the audio and video signals of the content are stored in the temporary storage unit 22 for the section, and the voice / non-speech discrimination unit 11 discriminates between the speech and non-speech sections of the audio signal in the section (Step S402). Note that the determination result determined in step S402 and the start and end times of the speech section are also stored in the temporary storage unit 22.

  Following step S402, the audio content rate calculation unit 13 calculates the audio content rate of the section (step S403). Based on the voice content rate calculated in step S403 and the section companding ratio calculated by the companding ratio calculating unit 20, the speed ratio condition setting unit 14 calculates the average speed ratio of the speech section and the average speed ratio of the non-speech section. , And the termination speed ratio of the voice section are calculated (steps S404 and S405). This process is the same as that of the first embodiment. Next, in step S406, the speed ratio condition setting unit 14 compares the terminal speed ratio Vend1 calculated in step S405 with the upper limit speed ratio Vend2 of the terminal speed ratio designated by the user or set in advance in the apparatus. When it is determined in step S406 that the terminal speed ratio Vend1 is larger than the upper limit speed ratio Vend2, the speed ratio condition setting unit 14 corrects the terminal speed ratio Vend1 to the upper limit speed ratio Vend2 (step S407). If it is determined in step S406 that the terminal speed ratio Vend1 is smaller than the upper limit speed ratio Vend2, the process proceeds to step S408.

  Here, the section audio content rate is a value in a section constituting a part of the content. Therefore, if there is a section where speech sections are locally concentrated, the value of the section speech content rate may increase locally. By performing the processes of steps S406 and S407 described above, it is possible to prevent the section voice content rate from increasing and the termination speed ratio of the voice section from becoming too large. Note that the statistical amount calculation unit 21 described in the third embodiment may calculate the long-term voice content rate and correct the upper limit speed ratio.

  After step S407, the speech segment length calculation unit 15 receives the start / end time of the speech segment and calculates the speech segment length (step S408). The speed ratio determination unit 16 refers to the start / end time of the speech section stored in the temporary storage unit 22 and determines whether or not the speech section is a predetermined unit time in order from the start end of the section (step S409). ). When it is determined that the voice segment is determined, the speed ratio determination unit 16 calculates the elapsed rate in the voice segment based on the start / end time of the voice segment and the voice segment length calculated in step S408 (step S410).

  Following step S410, the speed ratio determination unit 16 determines the conversion speed ratio of the voice section from the elapsed rate of the voice section (step S411). The processing in step S411 is the same as that in the first embodiment. If it is determined in step S409 that it is a non-speech segment, the speed ratio determination unit 16 determines the average speed ratio of the non-speech segment set by the speed ratio condition setting unit 14 as the conversion rate ratio from the start to the end of the non-speech segment. (Step S412).

  Subsequent to steps S411 and S412, the speed ratio determination unit 16 determines whether or not the conversion speed ratio has been calculated up to the end of the section (step S413). If it is not the end, the process returns to step S409. In this way, the processing from step S409 to step S413 is repeated in the speed ratio determination unit 16 until the conversion speed ratio up to the end of the section is calculated. If it is determined in step S413 that the conversion speed ratio has been calculated up to the end of the section, the speed conversion unit 17 converts the speed of the audio signal according to the conversion speed ratio, and starts reproducing the audio signal after the speed conversion ( Step S414). The speed ratio determination unit 16 determines whether or not the content has been played up to the end time of the content to be speed converted (step S415). If it is not the end time, the audio signal for the next section is stored in the temporary storage unit 22, and the process returns to step S402. If it is determined in step S415 that the content has been played back until the end time of the content, the input device (not shown) receives an instruction as to whether or not to end the processing of this device (step S416). When the user performs the speed conversion process for other contents (No in step S416), the process returns to step S401.

  As described above, according to the audio reproduction device according to the present embodiment, speed conversion can be performed in units of sections. Here, it is assumed that, for example, the user has recorded a program being broadcast from the beginning, but can be viewed from the middle of the program broadcast. At this time, since the user missed the beginning of the program, when the user wants to view the beginning with speed conversion, the user designates the start time of the speed conversion process as the beginning time. Note that the end point of the speed conversion process is the time when the latest recording is performed. That is, a section from the beginning to the time when the latest recording is made becomes one section. As a result, the user can perform viewing after speed conversion from the beginning to the time when the latest recording is performed, and thereafter can continue viewing by normal reproduction. As described above, according to the audio reproduction device according to the present embodiment, the speed conversion process is performed in section units. Therefore, even when content is being recorded, the speed conversion process can be performed without waiting for the end of the entire recording. it can.

  Moreover, according to the audio reproduction device according to the present embodiment, the temporary storage unit 22 is provided, whereby an actual measurement value of the audio content rate can be calculated, and the second and third implementations using the predicted value as the audio content rate. The speed conversion can be performed with a more optimal speed ratio as compared with the form. Further, by providing the temporary storage unit 22, it is possible to calculate an actual measurement value of the voice section length. As long as the actual measurement value is used for the voice section length, the termination speed ratio does not increase excessively due to the unknown voice section length, and is more optimal than the second and third embodiments using the predicted value as the voice section length. Speed conversion can be performed with a simple speed ratio.

(Fifth embodiment)
With reference to FIG. 29, an audio reproducing apparatus according to the fifth embodiment of the present invention will be described. FIG. 29 is a block diagram illustrating a configuration example of an audio reproduction device according to the fifth embodiment. In FIG. 29, the sound reproducing apparatus includes a sound discriminating unit 23, a storage unit 12, a sound content rate calculating unit 13, a speed ratio condition setting unit 14, a sound section length calculating unit 15, a speed ratio determining unit 16, and a speed converting unit 17. , And a specific event content rate calculation unit 24.

  In the first embodiment described above, the speed ratio between the voice section and the non-voice section is calculated. However, in this embodiment, it is possible to further calculate individual speed ratios for the specific event section included in the content. An audio playback device will be described. In addition, the audio playback device according to the present embodiment is different from the audio playback device according to the first embodiment in that a sound discrimination unit 23 is provided instead of the audio non-speech discrimination unit 11 and a specific event content rate calculation unit 24 is provided. Furthermore, it differs greatly in the point provided.

  The sound discriminating unit 23 receives an audio signal as an input, and discriminates a specific event section including a specific event sound, a voice section other than the specific event section, and a non-voice section. The specific event sound may be a sound from an individual sound source, or may be a group of sounds from a plurality of sound sources. Examples of the sound from the individual sound source include sound from the speaker A, sound from the musical instrument B, and specific sound from the device C. Examples of a set of sounds from a plurality of sound sources include a set of sounds from a plurality of speakers, music, noise, and the like. Further, the number of specific event sounds is not limited to one, and may be plural. When there are a plurality of specific event sounds, the sound discriminating unit 23 discriminates a plurality of specific event sections, a voice section other than the specific event sections, and a non-voice section by using the audio signal as an input. Further, when the specific event sound is a voice of speaker A, speaker B, or the like, the voice section determined by the sound determination unit 23 means a voice section other than the specific event section. In the following description, the specific event sound is assumed to be music.

The sound discriminating unit 23 receives the audio signal as an input, and discriminates a music section that is a specific event section, a voice section other than the music section, and a non-voice section. As a method for discriminating these sections, for example, “audio indexing from MPEG encoded data” <Yasuyuki Nakajima, Rikuyo, Masaru Sugano, Hiromasa Yanagihara, Ikuo Yoneyama (KDDI R & D Laboratories) 2000, Science D-II
, Vol. J83-D-II, no. 5, pp. 1361-1371> are known methods. In this method, the audio signal is first classified into a sound part and a soundless part. Further, the sounded part is further classified into three categories of voice, music, and cheer using Bayesian estimation. With such a method, the sound discriminating unit 23 discriminates a music segment, a voice segment other than the music segment, and a non-speech segment. Note that the cheers are included in a non-voice section other than the music section. The discrimination result discriminated by the sound discriminating unit 23, the start / end time of the voice section, and the start / end time of the music section are stored in the storage unit 12.

  The specific event content rate calculation unit 24 calculates the content rate of the specific event from the start and end times of the specific event section accumulated in the accumulation unit 12. The specific event content rate indicates a ratio of a specific event section (here, a music section) included in the audio signal of the content. In the following description, the specific event content rate is described in other words as the music content rate. Specifically, the music content rate is obtained by dividing the sum of music section lengths included in an audio signal for a predetermined time by the predetermined time. Here, it is assumed that the sum of music section lengths included in the entire content is divided by the content length.

  The speed ratio condition setting unit 14 first calculates an average speed ratio (hereinafter referred to as a music speed ratio) of a music section from the target companding ratio and the music content rate calculated by the specific event content rate calculation unit 24. The target companding ratio may be set by the user or may be set in advance in the apparatus. Specifically, the speed ratio condition setting unit 14 calculates the music speed ratio based on a table or a corresponding function that indicates the correspondence between the music content ratio and the music speed ratio that differ depending on the target companding ratio. It is assumed that this table and the corresponding function are prepared in advance. FIG. 30 is a diagram showing an example of a table showing the correspondence between the music content rate and the music speed ratio. The table shown in FIG. 30 shows the correspondence when the target companding ratio is 0.5.

例えば目標圧伸比が０．５、音楽含有率が１０％の場合、図３０により、音楽速度比は１倍速となる。したがってこの場合、式（１９）にＥ＝０．５、Ｓｍ＝０．１、Ｆ＝１を代入すると、Ｇ＝２．２５となる。 Based on the calculated music speed ratio, the speed ratio condition setting unit 14 calculates a companding ratio for a speech segment other than the music segment and a non-speech segment. Here, assuming that the music content rate is Sm, the music speed ratio is F, the target companding ratio is E, and the average speed ratio of speech sections other than the music section and non-speech sections is G, the average speed ratio G is given by Equation (19). Become.

For example, when the target companding ratio is 0.5 and the music content rate is 10%, the music speed ratio becomes 1 × speed as shown in FIG. Therefore, in this case, if E = 0.5, Sm = 0.1, and F = 1 are substituted into the equation (19), G = 2.25.

  The draw ratio can be expressed as the reciprocal of the speed ratio. Since the average speed ratio G is 2.25, the companding ratio of the voice segment other than the music segment and the non-speech segment is 0.44. Therefore, if this companding ratio (0.44) is set as the target companding ratio for the speech segment and the non-speech segment, the average speed ratio of the speech segment and the non-speech segment in the same manner as in the first embodiment. The average speed ratio and the end speed ratio of the voice section can be calculated. The speed ratio calculation distribution used by the speed ratio condition setting unit 14 may be set according to the music content rate.

  Hereinafter, with reference to FIG. 31, the process of the audio reproduction device according to the fifth embodiment will be described. FIG. 31 is a flowchart showing a flow of processing of the audio reproduction device according to the fifth embodiment.

  First, when a user gives an instruction to record content on an input device (not shown), an audio signal and a video signal of the content are stored in the storage unit 12. At this time, the sound discriminating unit 23 discriminates a music segment, a voice segment other than the music segment, and a non-speech segment (step S501). Note that the determination result determined in step S501, the start / end time of the voice section, and the start / end time of the music section are also stored in the storage unit 12.

  Next to step S501, it is determined whether or not the user has instructed the user to reproduce the desired content (step S502). When there is an instruction from the user (Yes in step S502), the audio content rate calculation unit 13 calculates the audio content rate of the instructed content (step S503). Further, the specific event content rate calculation unit 24 calculates the music content rate of the instructed content (step S504).

  After step S504, the speed ratio condition setting unit 14 calculates a music speed ratio from the target companding ratio and the music content rate calculated by the specific event content rate calculation unit 24 (step S505). Based on the calculated music speed ratio, the speed ratio condition setting unit 14 calculates the companding ratio of the speech section other than the music section and the non-speech section using Expression (19). Then, using the calculated companding ratio, the average speed ratio of the voice section, the average speed ratio of the non-voice section, and the terminal speed ratio of the voice section are set as speed ratio conditions (step S506). The voice segment length calculation unit 15 receives the start / end time of the voice segment and calculates the voice segment length (step S507).

  After step S507, the speed ratio determination unit 16 refers to the start / end times of the music sections stored in the storage unit 12, and determines whether the music section is a music section every predetermined unit time in order from the start end of the content. Judgment is made (step S508). If it is determined as a music section, the speed ratio determination unit 16 determines the music speed ratio calculated in step S505 as a conversion speed ratio (step S509). That is, the conversion speed ratio from the beginning to the end of the music section is constant at the music speed ratio.

  If it is determined in step S508 that it is not a music section, the speed ratio determining unit 16 refers to the start / end time of the voice section and determines whether it is a voice section (step S510). When it is determined that the voice segment is determined, the speed ratio determination unit 16 determines the voice based on the average speed ratio of the voice segment set in step S506, the start / end time of the voice segment, and the voice segment length calculated in step S507. The elapsed ratio in the section is calculated (step S511). The speed ratio determining unit 16 determines the conversion speed ratio of the voice segment from the elapsed rate of the voice segment (step S512). When determining that it is not a speech section, the speed ratio determination unit 16 determines the average speed ratio of the non-speech section set in step S506 as the conversion speed ratio (step S513). That is, the conversion speed ratio from the beginning to the end of the non-speech section is constant at the average speed ratio of the non-speech section. Note that the processing in steps S511 to S513 is the same as that in the first embodiment.

  After steps S509, S512, and S513, the speed ratio determination unit 16 determines whether the conversion speed ratio has been calculated up to the end of the content (step S514). If it is not the end, the process returns to step S508. In this way, the processing from step S508 to S514 is repeated in the speed ratio determination unit 16 until the conversion speed ratio up to the end of the content is calculated. If it is determined in step S514 that the conversion speed ratio has been calculated up to the end of the content, the speed conversion unit 17 converts the speed of the audio signal according to the conversion speed ratio, and starts playing the audio signal after the speed conversion ( Step S515). The input device (not shown) receives an instruction as to whether or not to end the processing of this device (step S516). When the user performs the speed conversion process for other contents (No in step S516), the process returns to step S502.

  As described above, according to the audio reproduction device according to the present embodiment, the specific event section is calculated when the specific event section is calculated and the speed ratio of the specific event section is set to view a music program or the like. The music section can be played back at a slower speed than the other voice sections and non-voice sections. Thereby, it is possible to perform reproduction with emphasis on music in the speed conversion process. In addition, when the specific event sound is not music but the voice of a certain speaker A appearing in the content, the voice of the speaker A is emphasized and the speaker is slower than the voice section other than the specific event section. A voice is subjected to speed conversion processing. For example, when it is desired to confirm the content of the speech of the speaker A with respect to the content viewed many times, it is useful to reproduce only the content of the speech of the speaker A at a slow speed. Also, when recording for a long time like a security camera, it is possible to identify the noise part as a specific event sound and play that part at high speed to reduce the time to see redundant scenes. . In this way, by setting an individual speed for a specific event sound, it can be played back at a slow speed to promote emphasis on that part, or it can be played back at a fast speed to reduce redundant parts. Speed setting according to the application.

  Note that the audio reproduction apparatuses according to the first to fifth embodiments described above may be realized by causing a general computer system 50 to execute an audio reproduction program. FIG. 32 is a block diagram illustrating a configuration example in which the audio reproduction device is realized by the computer system 50.

  32, the computer system 50 includes a CPU 51, a memory 52, a hard disk 53, a disk drive device 54, a monitor 55, a speaker 56, and an input device 57. CPU51 implement | achieves the same function as each structure part of the audio | voice reproduction apparatus based on 1st-5th Embodiment other than the storage part 12 and the temporary storage part 22 mentioned above by running an audio | voice reproduction | regeneration program. The memory 52 and the hard disk 53 realize the same function as the storage unit 12 and the temporary storage unit 22 by executing a sound reproduction program.

  The disk drive device 54 reads out the sound reproduction program from the recording medium 58 in which the sound reproduction program for causing the computer system 50 to function as the sound reproduction device is stored. By installing the audio reproduction program in an arbitrary computer system 50, the computer system 50 can function as the above-described audio reproduction apparatus.

  The recording medium 58 is a recording medium in a format that can be read by the disk drive device 54 such as a flexible disk or an optical disk. The audio reproduction program may be installed in the computer system 50 in advance. The audio reproduction program may be provided through a telecommunication line such as the Internet. Moreover, the audio | voice reproduction | regeneration process may be a form processed entirely or partially by hardware.

  The monitor 55 displays a video signal recorded on the recording medium 58 read via the disk drive device 54, a video signal recorded on the hard disk 53, and the like. The speaker 56 converts the audio signal recorded on the recording medium 58 read via the disk drive device 54, the audio signal recorded on the hard disk 53, and the audio signal after the speed conversion process into sound and reproduces it. The input device 57 is composed of, for example, a keyboard and a mouse, and receives an input of a target companding ratio.

  As described above, the sound reproduction apparatuses according to the first to fifth embodiments described above are realized by causing a general computer system 50 to execute a sound reproduction program.

  In addition, the audio reproduction apparatuses according to the first to fifth embodiments described above may be realized by a single chip using an integrated circuit such as an LSI or a dedicated signal processing circuit. Moreover, the audio | voice reproduction apparatus which concerns on the 1st-5th embodiment mentioned above may be implement | achieved by what each comprised the thing equivalent to the function of each structure part which comprises an audio | voice reproduction apparatus as a chip | tip. Here, although LSI is used, it may be called IC, system LSI, super LSI, or ultra LSI depending on the degree of integration. Further, the method of circuit integration is not limited to LSI's, and implementation using dedicated circuitry or general purpose processors is also possible. An FPGA (Field Programmable Gate Array) that can be programmed after manufacturing the LSI or a reconfigurable processor that can reconfigure the connection and setting of circuit cells inside the LSI may be used. Further, if integrated circuit technology comes out to replace LSI's as a result of the advancement of semiconductor technology or a derivative other technology, it is naturally also possible to carry out function block integration using this technology.

  The audio reproducing apparatus according to the present invention is a device for viewing AV contents such as a hard disk recorder and a DVD recorder, a personal computer and a mobile phone capable of performing an appropriate speed conversion according to an input audio signal while achieving a target time. This is useful for applications that run on mobile devices such as telephones. Also, not only for viewing purposes, but also for use in learning content playback systems, etc., for easy understanding of content, and for easy understanding of long-term content such as video taken with a security camera Etc. are also useful.

1 is a block diagram showing a configuration example of an audio reproduction device according to a first embodiment. Diagram showing audio content by genre A diagram showing the average and standard deviation of the audio content of each genre Diagram showing five types of calculation patterns Diagram showing an example of speed ratio calculation distribution The figure which shows the target companding ratio when the voice content rate is 0.5, the average speed ratio of the voice section, and the average speed ratio of the non-voice section Schematic showing the speed ratio change between voice and non-voice segments The flowchart which shows the flow of a process of the audio | voice reproduction apparatus which concerns on 1st Embodiment. The figure which showed the voice section length and the frequency which are included in the news program The figure which showed the voice section length and the frequency included in the baseball program Diagram showing change in companding ratio in voice interval The figure which shows the case where the conversion speed ratio of 2 steps is calculated The figure which shows the example of speed ratio calculation distribution about the genre where there are many still images such as documentary In the speed ratio calculation distribution shown in FIG. 13, the target companding ratio when the voice content rate is 0.5, the average speed ratio of the voice section, and the average speed ratio of the non-voice section The figure which shows the example of the speed ratio calculation distribution about the genre with many scenes with a lot of movements, such as sports In the speed ratio calculation distribution shown in FIG. 15, the target companding ratio when the voice content rate is 0.5, the average speed ratio of the voice section, and the average speed ratio of the non-voice section The block diagram which shows the structural example of the audio | voice reproduction apparatus which concerns on 2nd Embodiment. The figure which shows typically the method of calculating | requiring the prediction audio | voice content rate Y (z) shown by Formula (11). The figure which shows an example of the calculation result of the audio | voice content rate X (z) and the prediction audio | voice content rate Y (z) The figure which shows the audio | voice area length estimated based on Formula (14) The flowchart which shows the flow of a process of the audio | voice reproduction apparatus which concerns on 2nd Embodiment. The block diagram which shows the structural example of the audio | voice reproduction apparatus which concerns on 3rd Embodiment. The figure which shows the time change of long-term audio content rate for every contents The figure which shows the prediction voice content rate for every contents The flowchart which shows the process of the speed ratio condition setting part 14 which concerns on 3rd Embodiment. The figure which shows distribution of the actual measurement value of the voice section length, the actual measurement value of the last voice section length, and the predicted voice section length The block diagram which shows the structural example of the audio | voice reproduction apparatus which concerns on 4th Embodiment. The flowchart which shows the flow of a process of the audio | voice reproduction apparatus which concerns on 4th Embodiment. The block diagram which shows the structural example of the audio | voice reproduction apparatus which concerns on 5th Embodiment. The figure which shows the example of the table which showed a response | compatibility with a music content rate and a music speed ratio The flowchart which shows the flow of a process of the audio | voice reproduction apparatus which concerns on 5th Embodiment. FIG. 2 is a block diagram showing a configuration example in which an audio reproduction device is realized by a computer system 50 Block diagram showing the configuration of a conventional audio playback device

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Voice non-voice discrimination | determination part 12 Accumulation part 13 Voice content rate calculation part 14 Speed ratio condition setting part 15 Voice interval length calculation part 16 Speed ratio determination part 17 Speed conversion part 18 Voice content rate prediction part 19 Voice area length prediction part 20 Pressure Elongation ratio calculation unit 21 Statistics calculation unit 22 Temporary storage unit 23 Sound discrimination unit 24 Specific event content rate calculation unit 50 Computer system 51 CPU
52 Memory 53 Hard Disk 54 Disk Drive Device 55 Monitor 56 Speaker 57 Input Device

Claims (24)

An audio playback device that plays back a predetermined playback time by changing the playback speed of an audio signal of input content,
A discriminating means for discriminating a voice section including voice and a non-voice section not containing voice with respect to the audio signal;
A voice content rate calculating unit that calculates a voice content rate indicating a ratio of a voice section included in the audio signal based on the determination result determined by the determination unit;
Speeds for calculating the speed ratios of the voice and non-voice sections to the playback speed set in advance in the audio signal based on the voice content rate so that the playback time of the audio signal becomes the predetermined playback time. A ratio calculating means;
An audio playback device comprising: the audio signal as input, and speed conversion means for converting the playback speeds of a voice section and a non-voice section included in the audio signal based on the speed ratio. The speed ratio calculating means includes
An audio that indicates a companding ratio indicating a ratio of companding a reproduction time reproduced at a reproduction speed set in advance to the audio signal to the predetermined reproduction time, an audio content ratio, and an average speed ratio of the audio section. Using the correspondence information indicating the correspondence with the calculation method of the average speed ratio and the non-voice average speed ratio indicating the average speed ratio of the non-voice section, the speed is calculated by calculating the voice average speed ratio and the non-voice average speed ratio, respectively. Speed ratio condition setting means for setting as a ratio condition;
The speed ratio in each section into which the voice section is subdivided is determined as a speed ratio based on the voice average speed ratio, and the speed ratio in each section into which the non-voice section is subdivided is used as the non-voice average speed ratio. The audio reproduction device according to claim 1, further comprising: a speed ratio determining unit that determines a speed ratio based on the speed ratio and calculates a speed ratio between the voice section and the non-voice section. The voice reproduction device further includes voice section length calculating means for calculating each time from the start time to the end time of each voice section determined by the determining means as a voice section length.
The speed ratio condition setting means further sets, as a speed ratio condition, an end speed ratio at the end time of the voice section according to the voice average speed ratio,
The speed ratio determining means calculates a speed ratio in each section of the voice section based on the voice average speed ratio, the voice section length, and the end speed ratio for each voice section discriminated by the discrimination means. The audio reproduction device according to claim 2, wherein the audio reproduction device is determined.   The speed ratio determining unit is configured to determine, based on an elapsed ratio obtained by dividing a time elapsed from the start time of the voice segment by the voice segment length, for each voice segment determined by the determination unit. The audio reproduction device according to claim 3, wherein a speed ratio in each section is determined.   3. The audio reproduction device according to claim 2, wherein the speed ratio determination unit determines a speed ratio in each section of the audio section so that the reproduction speed increases as time elapses from the start time of the audio section.   The speed ratio condition setting means is correspondence information including at least one type of calculation method of the voice average speed ratio and the non-voice average speed ratio, and shows different correspondence depending on the type of content constituted by the audio signal. The audio reproduction device according to claim 2, wherein the audio average speed ratio and the non-audio average speed ratio are calculated using correspondence information.   The speed ratio condition setting means creates the correspondence information so that the voice average speed ratio and the non-voice average speed ratio are within a range specified by a user, and uses the correspondence information to generate the voice average speed ratio and The audio reproduction device according to claim 2, wherein the non-audio average speed ratio is calculated.   The said correspondence information is information which shows the correspondence from which the said audio | voice content rate differs from the calculation method of the said audio | voice average speed ratio and a non-voice average speed ratio according to the magnitude | size of the said audio | voice content rate. Audio playback device.   The speed ratio condition setting means is correspondence information including at least one kind of calculation method of the voice average speed ratio and the non-voice average speed ratio, and uses correspondence information indicating different correspondence depending on a user's purpose of use, The sound reproducing apparatus according to claim 2, wherein the sound average speed ratio and the non-voice average speed ratio are calculated.   The said correspondence information is information which shows the correspondence from which the said companding ratio differs from the calculation method of the said voice average speed ratio and the non-voice average speed ratio according to the magnitude | size of the said companding ratio. Audio playback device. An accumulator that accumulates in advance the audio signal that forms the entire content and the determination result determined by the determining unit with respect to the audio signal that forms the entire content;
The audio content rate calculating unit calculates an audio content rate indicating a ratio of audio segments included in an audio signal constituting the entire content based on a determination result stored in the storage unit in advance. The audio reproducing device described.   The audio reproduction according to claim 1, wherein the voice content rate calculating unit sequentially calculates a voice content rate used when the speed ratio calculating unit calculates based on a discrimination result discriminated in the past by the discrimination unit. apparatus. The voice content rate calculating means is used when the speed ratio calculating means calculates based on a determination result determined in the past for a first predetermined time from when the speed ratio calculating means calculates. Are sequentially calculated for each second predetermined time that is equal to or shorter than the first predetermined time,
The audio playback device is configured to determine the amount of data input to the speed conversion unit, the amount of data output from the speed conversion unit, and a playback time for playback at a playback speed set in advance in the audio signal. A drawing ratio calculating means for sequentially calculating a drawing ratio for each second predetermined time based on a drawing ratio indicating a ratio of drawing in time;
The speed ratio calculating means includes
The companding ratio for each second predetermined time, the voice content ratio for each second predetermined time, the voice average speed ratio indicating the average speed ratio of the voice section, and the average speed ratio of the non-voice section Using the correspondence information indicating the correspondence with the calculation method of the non-voice average speed ratio, the voice average speed ratio and the non-voice average speed ratio are calculated respectively, and the calculated voice average speed ratio and the non-voice average speed ratio are set as speed. A speed ratio condition setting means for sequentially setting the ratio condition every second predetermined time;
With respect to the voice section and the non-voice section included in the second predetermined time, a speed ratio in each section into which the voice section is subdivided is determined as a speed ratio based on the voice average speed ratio; The speed ratio in each section obtained by subdividing the non-speech section is determined as a speed ratio based on the non-speech average speed ratio, and the speech is based on a speed ratio condition sequentially set every second predetermined time. Speed ratio determining means for calculating the speed ratio of each section and non-voice section,
The speed conversion means converts the playback speeds of a voice section and a non-voice section included in the audio signal based on the speed ratio of the voice section and the non-voice section calculated by the speed ratio determination means, respectively. The audio reproducing device according to 1. The sound reproducing device includes a statistic calculating unit that calculates a statistic for suppressing a change in the second predetermined time indicated by a speed ratio in each section of the sound section determined by the speed ratio determining unit. In addition,
The speed ratio condition setting means is based on the voice average speed and the voice average speed based on the statistic, the voice content rate at the second predetermined time, and the companding ratio at the second predetermined time. 14. The audio playback device according to claim 13, wherein an end speed ratio of a voice section is calculated, and the calculated average voice speed and end speed ratio are sequentially set every second predetermined time as a speed ratio condition.   The statistic calculation means includes audio included in the time from the start time of the content to the time when the speed ratio calculation means calculates based on the determination result from the start time of the content to the time when the speed ratio calculation means calculates. The audio reproduction device according to claim 14, wherein an audio content rate indicating a ratio of sections is calculated as the statistic. The voice reproduction device calculates a voice interval length that is a time from a start time to an end time of the voice interval used when the speed ratio calculation unit calculates based on a determination result determined in the past by the determination unit. Voice section length calculating means for sequentially calculating;
The speed ratio condition setting means sequentially sets an end speed ratio at an end time of a voice section according to the voice average speed ratio as a speed ratio condition, and further sequentially every second predetermined time,
The speed ratio determining means calculates a speed ratio in each section of the voice section based on the voice average speed ratio, the voice section length, and the end speed ratio for each voice section discriminated by the discrimination means. The audio reproduction device according to claim 13, wherein the audio reproduction device is determined.   The voice section length used when the speed ratio calculating means calculates is the voice section length that is equal to or longer than a predetermined section length among the voice section lengths calculated from the start and end times of the voice sections discriminated in the past by the discriminating means. The sound reproduction device according to claim 16, wherein the sound reproduction device is calculated based only on the sound.   The voice section length used when the speed ratio calculating unit calculates is calculated based on a maximum value of a voice section length calculated from a start time and an end time of a voice section determined in the past by the determining unit. Item 17. The audio playback device according to Item 16. A storage means for storing in advance the audio signal for a predetermined time and the determination result determined by the determination means for the audio signal for the predetermined time;
2. The voice content rate calculating unit calculates a voice content rate indicating a ratio of a voice section included in the audio signal for the predetermined time based on a determination result stored in the storage unit in advance. Audio playback device. The determining means determines, for the audio signal, a specific event section including a specific event sound, the voice section other than the specific event section, and the non-voice section,
The audio reproduction device further includes a specific event content rate calculating unit that calculates a specific event content rate indicating a ratio of the specific event section included in the audio signal based on a determination result determined by the determining unit.
The speed ratio calculating means calculates a speed ratio of the specific event section to a playback speed preset in the audio signal based on the specific event content rate, and the playback time of the audio signal is the predetermined playback time. So as to calculate the speed ratio of the voice section and the non-voice section other than the specific event section with respect to the playback speed set in advance in the audio signal based on the voice content rate, respectively.
The speed conversion means converts the playback speed between the specific event section included in the audio signal and the voice section and the non-voice section other than the specific event section based on the speed ratio. Audio playback device. An audio playback method of changing the playback speed of an audio signal of input content and playing it at a predetermined playback time,
A determination step for determining a speech section including speech and a non-speech section not including speech for the audio signal;
A voice content ratio calculating step for calculating a voice content ratio indicating a ratio of a voice section included in the audio signal based on the determination result determined in the determination step;
Speeds for calculating the speed ratios of the voice and non-voice sections to the playback speed set in advance in the audio signal based on the voice content rate so that the playback time of the audio signal becomes the predetermined playback time. A ratio calculating step;
And a speed conversion step of converting the playback speeds of a voice section and a non-speech section included in the audio signal based on the speed ratio. A program for causing a computer of an audio reproduction device to reproduce at a predetermined reproduction time by changing the reproduction speed of an audio signal of input content,
A determination step for determining a speech section including speech and a non-speech section not including speech for the audio signal;
A voice content ratio calculating step for calculating a voice content ratio indicating a ratio of a voice section included in the audio signal based on the determination result determined in the determination step;
Speeds for calculating the speed ratios of the voice and non-voice sections to the playback speed set in advance in the audio signal based on the voice content rate so that the playback time of the audio signal becomes the predetermined playback time. A ratio calculating step;
A program that causes the computer to execute a speed conversion step that receives the audio signal as input and converts the playback speed of a voice section and a non-voice section included in the audio signal based on the speed ratio.   A computer-readable recording medium on which the program according to claim 22 is recorded. An integrated circuit that plays back a predetermined playback time by changing the playback speed of an audio signal of input content,
A discriminating means for discriminating a voice section including voice and a non-voice section not containing voice with respect to the audio signal;
A voice content rate calculating unit that calculates a voice content rate indicating a ratio of a voice section included in the audio signal based on the determination result determined by the determination unit;
Speeds for calculating the speed ratios of the voice and non-voice sections to the playback speed set in advance in the audio signal based on the voice content rate so that the playback time of the audio signal becomes the predetermined playback time. A ratio calculating means;
An integrated circuit comprising: the audio signal as an input; and speed conversion means for converting the playback speed of a voice section and a non-voice section included in the audio signal based on the speed ratio.

Images