JPH08123483A - Speech time base conversion device - Google Patents

Abstract

PURPOSE: To perform conversion into the speed of a speech which is higher than a reproduction speed and lower than a recording speed while maintaining the interval at the time of recording when a sound signal is reproduced at a speed lower than that in the recording. CONSTITUTION: A speech which is reproduced at an M-fold speed (0<M<1) by a recording and reproduction part 101 is written in an input buffer 103 occasionally and a voiced/voiceless decision part 104 makes a voiced/voiceless decision. A time base control part 105 varies an expansion ratio for data read out of the input buffer 103 according to the decision result of a voiced section or voiceless section to expand the time base. The time-base expanded data written in an output buffer are properly converted by a D/A converter 111 into an analog signal in cycles T. A data remaining mount monitor part 109 measures the remaining amount of data which are not read out of the output buffer to the D/A converter 11 and still left and supplies the result to an expansion ratio control part 110. The expansion ratio control part 110 determines the expansion ratio of the voiced section and voiceless section independently according to a conversion rule corresponding to the remaining amount of data and supplied it to the time base control part 105.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice capable of arbitrarily extending the length of the time axis of the voice, which is required when the voice is played back at a low speed on a video tape recorder (VTR) or the like. The present invention relates to a time axis conversion device.

[0002]

2. Description of the Related Art Conventionally, there has been an audio time axis converter for reproducing an audio signal recorded at a certain speed at a speed different from the speed at the time of recording. For example, in a tape recorder, it is possible to adjust the running speed of the tape to speed up or slow down the playback speed. However, as the playback speed changes, the pitch also changes at the same time, making the content hard to hear. Therefore, there has been proposed an audio time base conversion device capable of changing the reproduction speed without changing the pitch.

A conventional speech time base converter will be described below with reference to the drawings. FIG. 4 is a block diagram showing the configuration of a conventional audio time base conversion device. In FIG. 4, 1 is a recording / reproducing unit for recording and reproducing acoustic signals, 2 is an A / D converter for converting reproduced analog signals into digital signals, 3 is a buffer memory for recording digital data, 4 Is a D / A converter, 5 is a control unit that controls writing of data to the buffer memory, and 6 is a read control unit that controls reading of data from the memory.

The operation of the speech time base conversion apparatus configured as described above will be described below. Here, a description will be given of an audio time axis conversion device that restores a pitch to a state at the time of recording and reproduces the audio signal when the audio signal is reproduced at a recording speed or less on a recording medium.

First, in the recording / reproducing unit 1, M at the time of recording
The acoustic signal is reproduced at double speed (0 <M <1). Here, the recording / reproducing unit is, for example, a VTR, a tape recorder, or the like. Next, the acoustic signal reproduced from the recording / reproducing unit 1 has a sampling period T / M inversely proportional to the reproduction speed,
It is converted into a digital signal by the A / D converter 2. Here, T is a sampling period that satisfies the sampling theorem for the acoustic signal at the time of recording, and in the case of an acoustic signal reproduced at M times speed, it is a period of 1 / M thereof. These A / D-converted digital signals are sequentially recorded in the buffer memory 3 at the cycle T / M by the write controller 5.
Here, if each digital signal recorded in the buffer memory 3 is read and reproduced at a cycle T, the pitch at the time of recording can be restored, but input signal data is insufficient to continuously output the output signal, and There is a blank section. For this reason, the read control unit 6 is provided with a section in which the digital signal stored in the buffer memory 3 is repeatedly read twice in a frame unit of several tens of msec so as to compensate for the insufficient data. The D / A converter 4 converts the digital signal read by the read control unit 6 into an analog signal at a sampling cycle T. By this series of processing, the voice time axis conversion can be realized without changing the pitch. For the technique of converting only velocity with constant pitch as described here, for example, "Tape recorder for compressing / expanding conversation time axis", Kosaka, Yokobori,
Fujita, Nikkei Electronics (1976.26)
It is explained in detail in.

FIG. 5 shows an example of processing in the case of 1/2 speed. (A) shows the data at the time of recording, and (b) shows the time position of the data accumulated in the buffer memory. The data string of (c) is reproduced by repeating each block of (b) twice at the sampling period T. The data string has the same pitch as the data string of (a), and the time axis has a double scale. It has become.

[0007]

SUMMARY OF THE INVENTION In the above-mentioned conventional example,
The pitch at the time of recording is maintained, the voice speed is the same as the reproduction speed of the recording medium, and it is converted to be slower than that at the time of recording. here,
When the reproduction speed of the recording medium is slowed down when it is desired to view the image information in detail slowly on a VTR or the like, when the conventional time axis conversion device is used, the audio speed is slowed down like the image. To go. Here, it is said that the range in which the human conversation speed can be changed without a sense of discomfort is about 0.75 to 1.5 times. Therefore, if the reproduction speed of the recording medium is too slow, mainly when it is desired to view the image information in detail slowly, the reproduced sound speed becomes unnecessarily slow and a sense of incongruity occurs, which makes it difficult to hear. Further, in order to avoid this, if an attempt is made to listen to audio at a speed higher than the current reproduction speed of the recording medium, audio signal data to be reproduced will be insufficient in terms of time. In this case, a lacking audio data section is regularly generated, and even if a means such as inserting silent data into this section is used, it becomes a discontinuous audio signal, resulting in an extremely unnatural reproduced sound.

The present invention is intended to solve the above-mentioned problems, and when reading from a recording medium at a reproduction speed lower than the recording speed, the speed of voice is not slowed down more than necessary and a discontinuity is not generated. It is an object of the present invention to provide a voice time axis conversion device capable of listening to a voice that is easy to hear.

[0009]

In order to solve the above-mentioned problems, an audio time axis converter according to a first aspect of the present invention is an audio signal stored in a recording medium which is M (0 <M <1) times as large as that at the time of recording. A reproducing unit for reading at a speed of, an A / D converter for converting an analog signal read by the reproducing unit into a digital signal, and an A / D converter
An input buffer that stores the output data of the D converter, a voiced / non-voiced determination unit that determines the voiced section and the voiceless section of the input signal, and an expansion ratio and a voiced rate for the voiced section based on the determination result of the voiced / voiceless determination section. A time axis control unit that performs time axis expansion by setting the expansion ratio for each section independently, an output buffer for recording the output data of the time axis control unit, and a data remaining of the data recorded in the output buffer. A data remaining amount monitoring unit that measures the amount, an expansion ratio control unit that determines an expansion ratio for time-axis conversion according to a predetermined conversion rule according to the data remaining amount obtained from the data remaining amount monitoring unit, and an output buffer And a D / A converter for converting the voice data recorded in the above into an analog signal.

According to another aspect of the present invention, there is provided an audio time base conversion device including an expansion ratio setting section for independently setting an expansion ratio for a silent section and an expansion ratio for a voiced section according to the remaining amount of data. .

According to a third aspect of the present invention, in the voice time axis converter, the silent section expansion ratio is set to 1 / M or more, and the voiced section expansion ratio is set to 1.0 to 1 / M, and the remaining data amount is set. An expansion ratio control unit that determines each expansion ratio based on the corresponding conversion rule is provided.

According to a fourth aspect of the present invention, there is provided a voice time base conversion apparatus which sets the voiced section expansion ratio to 1 when the remaining amount of data is less than a predetermined value.
Set to / M, otherwise set to the specified fixed value,
An expansion ratio control unit for determining the expansion ratio based on the conversion rule corresponding to the remaining amount of data in the silent section expansion ratio of 1 / M or more is provided.

[0013]

According to the above construction, based on the result of the judgment of the voiced / unvoiced state, the time axis compression is performed in which the expansion ratio of the voiceless section is made larger than that of the voiced section, and then the data is written in the buffer memory. At this time, the remaining amount of data recorded in the buffer memory is measured, the expansion ratio is increased as the remaining amount of the data decreases, and the expansion ratio is automatically adjusted even if the ratio of silent sections is small. By configuring so that sufficient data is constantly secured in the buffer memory, it is possible to play back with the playback speed in the voiced section kept as close as possible to the recording time as much as possible, and obtain slow playback sound that is easy to hear. be able to.

Further, according to the structure of claim 4, when the remaining amount of data, which is the number of data remaining in the buffer memory, is extremely small, the time axis extension is performed at the extension ratio of 1 / M even in the sound section. In this case, the expansion ratio in the silent section is adjusted based on the remaining amount of data. As a result, it is possible to obtain a natural reproduced sound with no discomfort, in which the output signal is not interrupted by emptying the buffer memory while reproducing the audio at a predetermined fixed value.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the arrangement of a speech time base converter according to the first embodiment of the present invention.

In FIG. 1, 101 is a recording / reproducing section for recording and reproducing an acoustic signal, and 102 is A / A for converting an analog signal reproduced by the recording / reproducing section 1 into a digital signal.
A D converter, 103 is an input buffer for temporarily recording the AD-converted acoustic signal, and 104 is a voice sound for determining whether the digital signal string read out from the input buffer is a voiced section or a silent section. A silence determination unit, 105 is a time axis control unit that performs time axis expansion processing on a signal read from the input buffer at a predetermined expansion ratio, and 106 is data reading from the input buffer and reading that controls the address thereof. A control unit, 107 is a write control unit that controls writing of data to the output buffer and its address, 108 is an output buffer that temporarily stores the data processed by the time axis control unit, and 109 is temporarily stored in the output buffer. A data remaining amount monitoring unit for monitoring the data size being monitored, 110 determines the expansion ratio of the time axis control unit according to the output of the data remaining amount monitoring unit. Stretch ratio control unit that, 111 is a D / A converter for converting the digital data recorded in the output buffer to an analog signal.

The operation of the speech time base conversion apparatus configured as described above will be described in detail below with reference to FIG.

First, from the recording / reproducing unit 101, M at the time of recording
The acoustic signal is read at a speed of (0 <M <1) times. Hereinafter, the speed means the relative speed to the recording speed. Here, assuming that the sampling period at the time of recording in the recording / reproducing unit 101 is T, the acoustic signal reproduced by the recording / reproducing unit 101 at M times speed is successively digital signal sequence at the sampling period T / M by the A / D converter 102. Is converted to
It is written in the input buffer 103. On the other hand, since the D / A converter 111 performs conversion into an analog signal at the same sampling period T as during recording, a signal 1 / M times the input signal per unit time is appropriately prepared in the output buffer. There must be. At that time, instead of applying the same proportion of time-axis expansion to the entire input signal, the time-axis expansion and contraction is performed in the silent section at a larger expansion ratio than in the voiced section so that the expansion ratio of the voiced section is lowered. Is the basic idea.

From the signal sequence read from the input buffer, the sound / silence judging section 104 judges whether the sample value sequence is a sound section or a sound section. This sound / silence judgment can be easily judged by a known technique. Based on this determination result, the time axis control unit 106 performs time axis expansion processing on the data read from the input buffer and outputs the data to the output buffer 108. In that case, in the silent section, time axis expansion is performed at the expansion ratio for the silent section,
In the voiced section, time axis extension is performed at the extension ratio for the voiced section. These expansion ratios are given set values by the expansion ratio control unit 110 based on the data remaining amount obtained by the data remaining amount monitoring unit 109.

The data remaining amount monitoring unit monitors the remaining amount of data that has been written in the output buffer but has not yet been output to the D / A converter 111. The expansion ratio for the sound and the expansion ratio for the silent section are determined. Therefore, the output buffer is prevented from becoming empty by adjusting the expansion ratio according to the amount of data accumulated in the output buffer.

The relationship between the remaining amount of data and the expansion ratio is shown in FIG.
It may be given by a linear function as shown in (a) or may be changed stepwise. In the example of FIG. 2A, the expansion rate is increased as the output buffer is closer to the empty state, so that the data can be easily stored in the output buffer. Especially, the expansion rate in the silent section is increased. This is to prevent the output buffer from becoming empty even if the expansion rate of the voiced section is lowered. In the example of FIG. 2B, the voiced section is reproduced at the extension rate 0, that is, at the same voice speed as at the time of recording, unless the remaining data amount becomes 0. In this case, in the state where the expansion rate of the voiced section is fixed to 1, if the voiced section continues, the remaining amount of data in the output buffer decreases rapidly. Therefore, the expansion rate of the silent section should be set to a large value. , It is easy to collect data in the output buffer. Although it is possible to increase the number of data so that the output buffer does not become empty by expanding the time axis, if the expansion ratio of a large value is given unnecessarily, it will exceed the capacity of the output buffer. Cannot be maintained. Therefore, the expansion ratio is kept small as the remaining amount of data increases.

In the following, the reproduction speed of the recording medium is set to 2 when recording.
The operation will be described by taking the case of ⅓ times (M = 2/3) as an example.

First, in the expansion ratio setting table of FIG. 2, when the remaining amount of data is 0, the expansion ratio of the sound section is set to 1.5,
The output buffer is prevented from becoming empty even if a voice is given to the input signal. When the remaining amount of data is almost equal to the output buffer capacity, the expansion ratio in the silent section needs to be suppressed to 1.5 or less.

FIG. 3 is a schematic diagram showing the processing state in the case of expanding the time axis in the silent section and the voiced section at different time axis expansion ratios with respect to the time axis. (B) shows the case where the sound is reproduced from the recording medium at the reproduction speed of 2/3 times with respect to the input signal at the time of recording of (a). Here, it is necessary to determine the expansion ratio of the silent section and the voiced section depending on the ratio of the silent section of the input signal. (C) and (d) show two examples with different ratios of silent sections. In the portion of the input signals 1 to 6, in the example of (c), 1, 2, and 3 are silent intervals, and 4,
Processing is performed when 5 and 6 are voiced sections.
In the example of (d), processing is performed when 1 and 2 are silent sections and 3, 4, 5, and 6 are voiced sections. In this example, since the expansion ratios of the voiced sections are both 1.0, the expansion ratio of the silent section is 2.0 in the example of (c) and 2.5 in the example of (d). If the ratio of the silent sections can be estimated in advance as in these examples, the output data can be continuously supplied to the D / A converter from the output buffer without excess or deficiency, so the expansion ratio may be fixed. However, the ratio of silence included varies depending on the type of source to be reproduced.
Therefore, by monitoring the amount of data stored in the output buffer, determining the expansion ratio based on that value, and absorbing the time deficiency of the output data in the output buffer,
Even for a voice in which the ratio of silence is unpredictable, it is possible to set the expansion ratios of the silent section and the voiced section independently.

For a description of the time axis conversion process, see, for example, "" Realization of high-quality voice speed conversion system by DSP ", Suzuki, Misaki, Institute of Electronics, Information and Communication Engineers, Speech Study Group materials
SP90-34, (1990.8.23) ”and the like.

By controlling the expansion ratio as described above, the expansion ratio for time-base expansion slightly changes depending on the ratio of the silent section, but it is less than the speed of the sound at the time of recording and faster than the reproduction speed of the recording medium. The voice signal can be heard at the speed of voice.

As described above, according to the present embodiment, the time axis expansion ratio is set independently for each of the voiced section and the silent section based on the remaining amount of data, and when the remaining amount of data is less than a predetermined fixed amount. By setting the expansion ratio of the voiced section to 1 / M and preventing the output signal from being interrupted, by controlling the expansion ratio that makes the voiced section as close as possible to the speed of voice during recording,
Even if the reproduction speed of the recording medium becomes slower, it is possible to obtain a reproduced sound that is easy to hear without a feeling of strangeness.

[0029]

As described above, according to the present invention, the recording speed M
The sound signal reproduced by a factor of 2 (0 <M <1) is used to determine the voiced section and the voiceless section by the voiced / voiceless determination section, and the time axis control section independently sets the voiced section and the voiceless section. It is expanded on the time axis at the expansion ratio and stored in the output buffer, and the expansion ratio of the silent section and the voiced section is determined according to a predetermined rule according to the remaining amount of data recorded in the output buffer, and the silent section expansion ratio is set to 1 / M or more, and the expansion ratio of voiced section is 1.0 or more 1 /
By setting it to be equal to or lower than M and changing each expansion ratio independently, the speed of the voice in the voiced section can be made faster than the reproduction speed.
Further, the voiced section expansion ratio is set to 1 / M when the remaining amount of data is less than or equal to a predetermined value, and is set to a designated fixed value in other cases, and the silent section expansion ratio is set to 1. By determining based on the conversion rule corresponding to the remaining amount of data in the range of / M or more, the voice speed can be reproduced at a constant value faster than the reproduction speed. Therefore, the speed of the voice in the voiced section can be output at a value closer to that at the time of recording. In addition, a data remaining amount monitoring unit is provided so that the silent expansion ratio and the sound expansion ratio, or only the silent expansion ratio can be adjusted according to the ratio of the silent section. Even if is given, the output signal can be reproduced without interruption.

As described above, according to the present invention, it is not necessary to listen to the audio signal at an unnecessarily slow audio speed even if the reproduction speed of the recording medium is slowed down because of the convenience of viewing the image signal slowly, which makes the viewer feel uncomfortable. It is possible to provide an audio time axis conversion device that enables slow-motion playback that is easy to hear.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an audio time base converter according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of an expansion ratio setting table according to this embodiment.

FIG. 3 is a schematic diagram of a time axis extension process of the present embodiment.

FIG. 4 is a block diagram of a conventional audio time axis converter.

FIG. 5 is a schematic diagram of conventional time-axis expansion processing.

[Explanation of symbols]

 Reference numeral 101 recording / playback unit 102 A / D converter 103 input buffer 104 voiced / non-voiced determination unit 105 time axis control unit 106 read control unit 107 write control unit 108 output buffer 109 data remaining amount monitoring unit 110 expansion ratio control unit 111 D / A converter

Claims (4)

[Claims] 1. A reproducing section for reading an acoustic signal stored in a recording medium at a speed M (0 <M <1) times as high as that at the time of recording, and an analog signal read by the reproducing section is converted into a digital signal. A / D converter, an input buffer that stores the output data of the AD converter, a voiced / soundless determination unit that determines a voiced section and a silence section of the input signal, and a determination of the voiced / soundless determination section. According to the result, a time axis control unit that sets the expansion ratio for the silent section and the expansion ratio for the voiced section independently to expand the audio data in the input buffer on the time axis, and stores the output data of the time axis control unit. According to a predetermined rule according to the output buffer, a data remaining amount monitoring unit that measures the data remaining amount of the data stored in the output buffer, and the data remaining amount obtained from the data remaining amount monitoring unit Time axis conversion And stretch ratio control unit which determines the stretch ratio,
An audio time base conversion device comprising: a D / A converter for converting audio data recorded in the output buffer into an analog signal. 2. The audio time base conversion according to claim 1, wherein the expansion ratio control unit independently sets an expansion ratio for a silent section and an expansion ratio for a voiced section according to the remaining amount of data. apparatus. 3. The expansion ratio control unit sets the silent section expansion ratio to 1 / M.
3. The above-described setting, the sounding section expansion ratio is set to 1.0 or more and 1 / M or less, and each expansion ratio is determined based on the conversion rule corresponding to the remaining amount of data. Audio time base converter. 4. The expansion ratio control unit sets the expansion ratio of the voiced section to 1 / M when the remaining amount of data is equal to or less than a predetermined value, and sets it to a fixed value other than the specified value for the silent section. Stretch ratio 1 /
4. The audio time base converter according to claim 3, wherein the expansion ratio is determined based on a conversion rule corresponding to the remaining amount of data in a range of M or more.