JP2007101644A - Voice reproducing apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve easiness of listening by suppressing voice degradation in minimum, when voice data is reproduced with a speed faster than original voice. <P>SOLUTION: A reproduction speed rate is set and a compression data is deleted in a packet unit in a data thinning section 33. The input compression data is decompressed to the voice data in a compression data decompressing section 35. It is determined whether the input voice data is sound or non-sound and the voice data of a non-sound period is deleted and the voice data of a sound period is output, in a non-voice period deleting section 37. A pitch cycle of the input voice data is extracted in a pitch extracting section. The voice data is thinned with a predetermined rate with a pitch cycle unit extracted by a thinning section 45 and the remaining voice data is synthesized in a synthesizing section 47 and output. Each section is controlled by a voice processing control section 31. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an audio reproducing apparatus having a “early listening” function capable of quickly listening to recorded audio information with a reproduction time shorter than the recording time.

  2. Description of the Related Art Generally, devices that are used to record and record audio such as conferences and TV broadcasts and play them later are known. In addition, there is a demand for such a device to view in a shorter time than the normal playback time.

  In the past, some analog recording devices have realized such a fast listening function. However, simply increasing the number of rotations of the motor raises the pitch during playback and significantly degrades the sound quality. Yes, it was far from quality that met the user's requirements.

  In recent years, digital recording systems such as HDDs (Hard Disk Devices) and DVDs have adopted a digital recording system. As an advantage, the sound quality has been improved by signal processing circuits.

  Specifically, as shown in FIG. 6, the digital recording device converts an analog waveform audio signal into digital data by a signal processing circuit to generate LPCM (Liner Pulse Code Modulation) data and compresses the LPCM data. Recording on a recording medium such as an HDD or a DVD. Here, the following method is employed for high speed reproduction of audio data digitally recorded on the recording medium at twice the normal reproduction speed.

(1) First, when audio data recorded on a recording medium by the LPCM method is reproduced at a reproduction speed of 2 × speed, as shown in FIG. 7, if the audio data is output at a reproduction speed of 2 × speed, processing is performed. Has the advantage of being simplified.
(2) Secondly, when the audio data recorded on the recording medium by the LPCM method is reproduced at a normal speed of 1 × speed, the audio data is thinned out at a ratio of one to two as shown in FIG. Then, there is an advantage that the processing is simplified.
(3) Thirdly, when the compressed audio data is reproduced at a normal speed of 1 × speed, as shown in FIG. 9, when the packetized data sequence is thinned out at a rate of one in two, and output, There is an advantage that the pitch does not change and the processing is simplified.
Japanese Patent No.3081469

However, when the method described above is adopted,
(1) The first method has a disadvantage that the pitch (pitch) is doubled, and there is a problem that an analog conversion circuit capable of double-speed reproduction is also required in the configuration.
(2) The second method has a drawback that information is missing.
(3) The third method has a drawback that some information is missing.

  Therefore, a technology that can improve the ease of listening by minimizing the deterioration of the voice that is being heard and played quickly is desired.

  The present invention has been made in view of the above. For the purpose of the present invention, when audio data is reproduced at a higher speed than the original audio, the deterioration of the audio is minimized, and the ease of listening is improved. An object of the present invention is to provide an audio playback device that can perform the above.

  In order to solve the above-mentioned problem, the invention according to claim 1 is an audio processing device that generates audio data whose speech speed has been converted so that the reproduction speed becomes a predetermined reproduction magnification with respect to the input original audio data. The setting means for setting the reproduction magnification, the first decimation means for deleting the input compressed data in units of packets, and the compression for decompressing the compressed data output from the first decimation means to audio data Data decompression means, and silence interval deletion means for judging whether the voice data output from the compressed data decompression means is voiced / silent, deleting the voice data in the silent section, and outputting the voice data in the voiced section A pitch extracting means for extracting the pitch period of the voice data output from the silent section deleting means, and the voice data in a unit of the pitch period extracted by the pitch extracting means. A second decimation unit that outputs the audio data remaining after decimation, a synthesis unit that synthesizes and outputs the audio data remaining after decimation output from the second decimation unit, and the reproduction magnification is When set within the range of 1 to 1.5, the silent section deleting means controls the silent section to be deleted while leaving at least 150 msec of the silent section, and the second thinning means is 1 / After thinning out at a rate of 20, the remaining voice data is compared with the set playback time, and the voice data is further thinned out by the speech speed conversion process so that the playback of the voice data ends within the set playback time. And when the reproduction magnification is set within a range of 1.5 to 2, the silent section deleting means controls to delete the silent section leaving at least 100 msec of the silent section. The second thinning means compares the remaining audio data with the set reproduction time, and further thins out the data by the speech speed conversion process in the sound period so that the reproduction of the audio data is finished within the set reproduction time. And when the reproduction magnification α is set to be larger than 2, the first thinning means controls to delete the compressed data packet at a ratio of 1 to (reproduction magnification−2), and the silence The section deleting means controls to delete the silent section, and the second thinning means compares the remaining audio data with the set reproduction data, and controls so that the reproduction of the audio data ends within the set reproduction time. The gist of the present invention is that it includes a control means.

  According to the first aspect of the present invention, when the reproduction magnification is set within a range of 1 to 1.5, the silent section is controlled to be deleted while leaving at least 150 msec, and 1 After thinning out at the rate of / 20, the remaining audio data is compared with the set playback time, and the data is further thinned out by the speech speed conversion process in the voiced section so that the playback of the audio data is completed within the set playback time. When the playback magnification is set within a range of 1.5 to 2, the silent section is controlled to be deleted while leaving at least 100 msec, and the remaining audio data and the set playback time are set. Are controlled so that the data is further thinned out by the speech speed conversion process in the sound period so that the reproduction of the audio data is finished within the set reproduction time, and the reproduction magnification is set within a range of 2 to 3. In such a case, control is performed so that the compressed data is deleted at a rate of (α−2) per α, so that the data amount is twice the playback time, and the silent section is deleted, and the remaining Audio data and playback data are compared and control is performed so that playback of the audio data ends within the set playback time, thereby minimizing audio degradation when playing back audio data at a speed faster than the original audio. It is possible to improve the ease of hearing by limiting to the limit.

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

  FIG. 1 is a block diagram showing a configuration of a video / audio processing apparatus 11 applicable to an audio processing apparatus according to an embodiment of the present invention. The video / audio processing device 11 shown in FIG. 1 shows a part of the configuration of a hard disk (HDD) / DVD recorder device.

  The A / V encoder unit 15 compresses video and audio data input from the external input terminal 19 and outputs the compressed transport stream (TS) information to the A / V decoder unit 17 or the storage unit 23. .

  The storage unit 23 includes a hard disk drive (HDD) capable of storing compressed digital video and audio data, and temporarily stores transport stream (TS) information in a compressed state from the A / V encoder unit 15. Or output to the A / V decoder unit 17.

  The A / V decoder unit 17 separates the video data and the audio data from the transport stream (TS) information in which the video and the audio input from the A / V encoding unit 15 or the storage unit 23 are superimposed. The compressed video data and audio data are expanded into reproducible video data and audio data, the video data is output to an image display circuit (not shown), and the audio data is output to the audio processing unit 31.

  The control unit 27 has a microprocessor that executes a stored program, and has a desired function content corresponding to a user operation via a remote controller (not shown), a user input terminal 29, a key provided on the main body, or the like. In order to realize this desired function, a command is output to each unit and controlled.

  The voice processing unit 31 includes a data thinning unit 33, a compressed data decompression unit 35, a silent section deletion unit 37, a memory unit 39, a speech extraction unit 41 including a pitch extraction unit 43, a thinning unit 45, and a synthesis unit 47, and a memory unit. 49, a D / A converter 51, and an audio processing control unit 53.

  The data thinning unit 33 can perform the thinning process by deleting the compressed data, and can shorten the reproduction time according to the thinning ratio.

  The compressed data decompressing unit 35 decompresses the input data to LPCM data if the input data is compressed data such as AAC (Advanced Audio Coding) or AC3 (Audio Code number 3), and decompresses if the input data is in the LPCM system. The LPCM data is output as it is without performing any operation.

  The silent section deletion unit 37 determines whether the input voice data is voiced / silent, deletes the section determined to be silent or unnecessary, and stores the voice data of the voiced section as a memory. The voice data stored in the unit 39 and stored in the memory unit 39 are read out in the oldest order. The memory unit 39 records and holds the remaining valid audio data output from the silent section deleting unit 37.

  The speech speed conversion unit 41 detects the pitch (fundamental frequency) of the voice data, and performs a thinning process and a synthesis process to obtain a desired speech speed. That is, the pitch extraction unit 43 sequentially extracts the pitch period (basic period of vocal cord vibration) of the audio data. The thinning unit 45 performs thinning processing on the audio data at a ratio of 1 / n in units of the extracted pitch period. The synthesizing unit 47 executes a synthesizing process for synthesizing audio data in units of basic periods of the thinned audio sections. The memory unit 49 records and holds audio data during the thinning process.

  The D / A converter 51 converts the digital audio data output from the speech speed conversion unit 41 into an analog audio signal.

  The voice processing control unit 53 includes a microprocessor that executes a stored program, and performs a desired operation according to a user operation via a remote controller (not shown), a user input terminal 29, a key provided on the main body, or the like. The contents of the function are input, and in order to realize the desired function, a command is output to each unit and controlled.

  Next, the operation of the video / audio processing apparatus 11 will be described with reference to the flowcharts shown in FIG. 2 with reference to schematic diagrams showing the outline of each process shown in FIGS. This flowchart is stored in the voice processing control unit 53 as a program.

  First, in step S10, the voice processing control unit 53 determines whether or not the “early listening” key is turned on as an operation command input from a remote controller or an operation panel (not shown) via the control unit 27. This determination process is repeated until the “listen” key is turned on.

  When the “early listening” key is turned on, the process proceeds to step S20, where the audio processing control unit 53 sets the reproduction magnification α input from the remote controller or the operation panel (not shown) via the control unit 27 in the audio processing control unit 53. Once stored in the memory, the process proceeds to step S30. If there is no input of the reproduction magnification α, the stored value when it was previously input may be used again as the reproduction magnification α.

  Next, in step S30, the audio processing control unit 53 includes recording state information indicating that the audio data has the LPCM format in the information output from the A / V decoder unit 17 via the control unit 27. If the audio data is LPCM, the audio data is stored in the memory unit 39 and the process proceeds to step S70. If the audio data is not LPCM, the process proceeds to step S40.

  Here, in step S40, the audio processing control unit 53 determines whether or not the reproduction magnification α currently set in the internal memory is α ≦ 2.

  When the reproduction magnification α is α> 2 (in the case of NO), the process proceeds to step S50, where the audio processing control unit 53 controls the data thinning unit 33, and the amount of data is the reproduction time in units of input packet data. The packet data is deleted at a rate of (α-2) per α so as to be doubled, and the process proceeds to step S60.

  On the other hand, when the reproduction magnification α is α ≦ 2 (in the case of YES), the process proceeds to step S60, where the audio processing control unit 53 controls the compressed data decompressing unit 35, and the input data is compressed such as AAC or AC3. In the case of data, it is decompressed into LPCM data, and the audio data is stored in the memory unit 39.

  Next, in step S70, the voice processing control unit 53 controls the silent section deleting unit 37 to determine whether the voice data input to the silent section deleting unit 37 is voiced or silent. Output a silent number label with the data number of the silent section added. For example, as shown in FIG. 3, among input digital data, data Nos. 13 to 20 are silent sections.

  Next, in step S80, the sound processing control unit 53 compares where the set reproduction magnification α is in the divided range as follows.

(When 1 <α ≦ 1.5)
Here, with reference to the schematic diagram shown in FIG. 4, processing when the reproduction magnification α is 1 <α ≦ 1.5 will be described. Note that the normal playback time of recorded audio data is 60 minutes, and the following processing is performed assuming a viewing time of 40 minutes.

  In step S80, if the range of the set reproduction magnification α satisfies 1 <α ≦ 1.5, the audio processing control unit 53 proceeds to step S90.

  Next, in step S90, the silent section deleting unit 37 leaves a silent section of at least 150 msec as a pause between sentences for the LPCM data stored in the memory unit 39, and a silent number label ( LPCM data to which No. 13 to No. 20) is added is deleted and output to the speech speed conversion unit 41.

  The inter-sentence pause part refers to a silent period of a certain time or longer, and corresponds to breathing or “ma” between utterances. In general television broadcasting, 20 to 30% of audio data is a silent section, and in a conference or the like, about 50% of audio data is a silent section.

  Next, in step S130, the control unit 27 sets the expansion rate β of the voiced section to 0.95 for the speech speed conversion unit 41 via the voice processing control unit 53. Next, in step S140, the audio data (the audio data from which No. 1, No. 29 and No. 13 to No. 20 of No. 0 to No. 32 are deleted) is compared with the reproduction time (40 minutes), so as to be suitable for the reproduction time. Data is thinned out by speech speed conversion processing in a voiced section to shorten the playback time.

  That is, the voice data output from the silent section deletion unit 37 is stored in the memory unit 49 provided in the speech speed conversion unit 41. Next, the pitch extraction unit 43 sequentially extracts the pitch period of the audio data stored in the memory unit 49, and the decimation unit 45 decimates the audio data at a ratio of 1/20 in units of the extracted pitch period. I do. As a result, as shown in FIG. 4, data (No. 1, 29) is thinned out, and audio data that has not been thinned is output to an output buffer (not shown) provided in the synthesis unit 47. Further, the synthesizing unit 47 synthesizes the audio data read from the output buffer in units of the basic period of the audio section and outputs the synthesized data to the D / A converter 51. The D / A converter 51 converts the digital audio data output from the speech speed conversion unit 41 into an analog audio signal as in the reproduction waveform shown in FIG.

  Note that the expansion rate β is the original voice data and the processed voice data when the original voice data actually recorded is shortened / expanded on the time axis using the speech speed conversion process. The ratio. In this example, the expansion rate β of 0.95 means that the audio data is thinned out at a rate of 1/20 and the time is shortened. In general, when the expansion rate β is 0.95 to 1.05, a person is unaware of changes in the speech rate of speech.

  Next, in step S150, the voice processing control unit 53 determines whether or not the “stop” key is turned on as an operation command input from a remote controller or an operation panel (not shown) via the control unit 27, and the “stop” key. If is not ON, the process returns to step S30 and the processes of steps S30 to S150 are repeated. On the other hand, if the “stop” key is turned on, the process is terminated.

  Thus, when the reproduction magnification α is in the range of 1 <α ≦ 1.5, as shown in FIG. 4, the expansion rate β is 0.95 compared to the original audio data before audio processing. There is almost no change in the speech rate in the voice section, and in the voiced section, listening close to the normal playback becomes possible. In particular, although the voice content can be understood by the conventional technology, it is difficult to hear the speech quickly, but according to the present embodiment, there is an advantage that it is very easy to hear.

(If 1.5 <α ≦ 2)
In step S80, if the set range of the reproduction magnification α satisfies 1.5 <α ≦ 2, the audio processing control unit 53 proceeds to step S100. The normal playback time of recorded audio data is 60 minutes, and the following processing is performed assuming a viewing time of up to 30 minutes.

  Next, in step S100, the silent section deleting unit 37 leaves a silent section of at least 100 msec as a pause between sentences for the LPCM data stored in the memory unit 39, and a silent number label ( LPCM data to which No. 13 to No. 20) is added is deleted and output to the speech speed conversion unit 41.

  That is, the voice data output from the silent section deletion unit 37 is stored in the memory unit 49 provided in the speech speed conversion unit 41. Next, the pitch extraction unit 43 sequentially extracts the pitch period of the audio data stored in the memory unit 49, and the thinning unit 45 compares the audio data with the set reproduction time, and the reproduction of the audio data is within the set reproduction time. The data is thinned out by the speech speed conversion process in the voiced section so that the reproduction time is shortened. As a result, the data is thinned out and the sound data not thinned out is output to an output buffer (not shown) provided in the synthesis unit 47. Further, the synthesizing unit 47 synthesizes the audio data read from the output buffer in units of the basic period of the audio section and outputs the synthesized data to the D / A converter 51. The D / A converter 51 converts the digital audio data output from the speech speed conversion unit 41 into an analog audio signal.

  In this way, when the reproduction magnification α is in the range of 1.5 <α ≦ 2, the sound is more swift than the original sound before the sound processing, but the quality that can withstand listening is maintained. In particular, the conventional technique only has an auxiliary role as a voice guide and is not suitable for fast listening, whereas the present embodiment has an advantage that it is suitable for fast listening.

(If 2 <α)
With reference to the schematic diagram shown in FIG. 5, processing when the reproduction magnification α is 2 <α will be described. FIG. 5 shows the reproduction magnification α = 3.

  In step S50, for example, when the recording data is compressed sound source data, the audio processing control unit 53 uses (3-2) packets for every 3 packets in the data thinning unit 33 as shown in FIG. , Packets are thinned out at a constant interval of 1/3. As a result, the packet in which the packets (0-2) and (0-5) are thinned out from the data thinning unit 33 is input to the compressed data decompressing unit 35.

  In step S80, if the range of the set reproduction magnification α satisfies 2 <α, the audio processing control unit 53 proceeds to step S110.

  Next, in step S110, the silent section deleting unit 37 deletes the LPCM data to which the silent number label (numbers 15 to 20) is added as a silent section from the LPCM data stored in the memory unit 39, and the speech speed. The data is output to the conversion unit 41.

  Next, in step S140, the audio data (audio data in which numbers 10 to 20 and numbers 25 to 29 are deleted from the numbers 0 to 34) are compared with the set reproduction time (20 minutes), and reproduction of the audio data is set. Data is thinned out by speech speed conversion processing in a voiced section so as to end within the playback time, and the playback time is shortened.

  That is, the voice data output from the silent section deletion unit 37 is stored in the memory unit 49 provided in the speech speed conversion unit 41. Next, the pitch extraction unit 43 sequentially extracts the pitch period of the audio data stored in the memory unit 49, and the thinning unit 45 compares the audio data with the set reproduction time, and the reproduction of the audio data is within the set reproduction time. The data is thinned out by the speech speed conversion process in the voiced section so that the reproduction time is shortened. As a result, as shown in FIG. 5, the data (Nos. 2, 6, 9, 23, and 31) are thinned, and the audio data that has not been thinned is output to an output buffer (not shown) provided in the synthesis unit 47. . Further, the synthesizing unit 47 synthesizes the audio data read from the output buffer in units of the basic period of the audio section and outputs the synthesized data to the D / A converter 51. The D / A converter 51 converts the digital audio data output from the speech speed conversion unit 41 into an analog audio signal as in the reproduction waveform shown in FIG.

  In this way, when the reproduction magnification α is larger than 2, when 2 <α ≦ 3, as shown in FIG. 5, information is lost early compared to the original sound before sound processing. I understand the contents. In particular, it is difficult to hear with the conventional technique, but according to this embodiment, there is an advantage that it is audible and does not give any difficulty.

  Further, when the reproduction magnification α is in the range of 3 <α ≦ 4, information is lost more quickly than the original sound before the sound processing, but it is an auxiliary function as a sound guide. In particular, the present embodiment has an advantage that it is audible, whereas it cannot be heard by the conventional technique.

  Furthermore, when 4 <α, the voice is significantly deteriorated, but it may be useful as a voice guide. In particular, while the conventional technique cannot be heard and is troublesome to hear as noise, the present embodiment has the advantage of fulfilling its purpose depending on the use of the product or model.

  As described above, by using “speech speed conversion technology” that changes the speech speed on the time axis and “data thinning” in combination, and by changing each processing parameter according to the playback magnification α, it is possible to listen to the voice. Easiness can be improved.

It is a block diagram which shows the structure of the video / audio processing apparatus 11 applicable to the audio processing apparatus which concerns on one embodiment of this invention. 3 is a flowchart for explaining the operation of the video / audio processing apparatus 11; It is a schematic diagram which shows the audio | voice waveform with a silence area, and its audio | voice data. FIG. 6 is a schematic diagram of audio data and a reproduction waveform when a reproduction magnification α is 1 <α ≦ 1.5. It is a schematic diagram of audio data and a reproduction waveform when a reproduction magnification α is 2 <α ≦ 3. It is a schematic diagram of a speech waveform and speech data. It is a schematic diagram of audio data and reproduction waveform when LPCM data is reproduced at double speed. It is a schematic diagram of audio data and a reproduction waveform when LPCM data is thinned by half and then reproduced at 1 × speed. FIG. 4 is a schematic diagram of audio data and a reproduction waveform when compressed audio data is thinned out by 1/2 in packet units and then reproduced at 1 × speed.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Image | video audio | voice processing apparatus 15 A / V encoder part 17 A / V decoder part 23 Storage part 27 Control part 31 Voice processing circuit 33 Data thinning part 35 Compressed data decompression part 37 Silent section deletion part 39 Memory part 41 Speech speed conversion part 43 Pitch extraction unit 45 Thinning-out unit 47 Synthesis unit 49 Memory unit 51 D / A converter 53 Audio processing control unit (control means)

Claims (1)

A speech processing device that generates speech data whose speech speed has been converted so that a playback speed is a predetermined playback magnification with respect to input original speech data,
Setting means for setting the reproduction magnification;
A first thinning means for deleting input compressed data in units of packets;
Compressed data decompression means for decompressing the compressed data output from the first thinning means into audio data;
A silent section deleting means for determining whether the voice data output from the compressed data decompressing means is voiced / silent, deleting the voice data of the silent section and outputting the voice data of the voiced section;
Pitch extraction means for extracting the pitch period of the audio data output from the silent section deletion means;
A second decimation unit that outputs the remaining audio data after decimation of the audio data at a predetermined rate in units of the pitch period extracted by the pitch extraction unit;
Synthesizing means for synthesizing and outputting the audio data remaining after decimation output from the second decimation means;
When the reproduction magnification is set within a range of 1 to 1.5, the silent section deleting means controls to delete the silent section leaving at least 150 msec of the silent section, and the second thinning-out. After the means thins out at a rate of 1/20, the remaining voice data is compared with the set playback time, and the voice data is further converted by the speech speed conversion process in the voiced section so that the playback of the voice data ends within the set playback time Control to thin out
When the reproduction magnification is set within a range of 1.5 to 2, the silence interval deletion means controls to delete the silence interval while leaving the silence interval at least 100 msec, and the second decimation is performed. The means compares the remaining audio data with the set playback time, so that the audio data playback ends within the set playback time.
Furthermore, it controls to thin out the data by the speech speed conversion process
When the reproduction magnification is set in a range larger than 2, the first decimation means controls to delete the compressed data packet at a ratio of 1 to (reproduction magnification -2), and the silent section deletion means Control to delete the silent section,
In addition, the second thinning means compares the remaining audio data with the set reproduction time, and in the sound period, the data is further converted into a fixed period and in a sound period so that the reproduction of the audio data is finished within the set reproduction time. An audio processing apparatus comprising control means for performing control so as to be thinned out evenly.

Images