JP2008107706A - Speech speed conversion apparatus and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent that a wave shape of a voice signal does not become continuous after processing, without imposing an excessive processing load on a speech speed conversion apparatus, when speech speed is converted by applying time axis compression on the voice signal. <P>SOLUTION: The speech speed conversion apparatus comprises: a calculation means for calculating a value which indicates a time change rate of a logarithmic value of a sound volume which is expressed by the voice signal to be processed; a discrimination means for discriminating that speech speed conversion is inhibited, in frames in which a magnitude of the value calculated by the calculation means is more than a prescribed threshold value, and the prescribed number of frames following the frame, in the plurality of frames for constituting the voice signal; and a speech speed conversion means which outputs the voice signal as it is, for the frame which is discriminated that the speech speed conversion is inhibited, and meanwhile, which outputs the voice signal by inserting or deleting a wave shape so that it may become the specified speech speed, for the other frames. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a technique for performing time axis companding processing on an audio signal.

Various techniques have been proposed for performing time axis companding processing on an audio signal and appropriately adjusting the speech speed of the audio represented by the audio signal. For example, Non-Patent Document 1 calculates the autocorrelation of speech while changing the frame length, regards the frame length with the highest correlation as the cycle of the speech, and inserts or deletes waveforms in units of the cycle. An algorithm called PICOLA that performs speech speed conversion is disclosed.
Naotaka Morita and Fumada Itakura, “Expansion and compression of speech over time using pointer movement control (PICOLA) and its evaluation”, Proc. Of the Acoustical Society of Japan, p.149-150, October 1986 Moon

  However, if the technique disclosed in Non-Patent Document 1 is simply applied, there may be a problem in hearing in a portion where the volume change is large or a consonant portion of a plosive sound. Specifically, if a waveform is inserted or deleted at a portion where the volume change is large, the change in the volume becomes discontinuous at that portion, and there may be an audible noise “bumpy”. In addition, if a waveform is inserted or deleted at the consonant part of the plosive sound, the consonant may be heard multiple times.

  As a measure for avoiding the occurrence of waveform discontinuity as described above, speech recognition processing is performed on the speech signal to be processed, a location where waveform discontinuity such as a consonant of a plosive is likely to occur is specified in advance, It may be possible to exclude it from the target of the time axis companding process. However, the execution of the speech recognition process generally requires a large amount of hardware resources (in other words, a high processing load), and there is a problem that it is difficult to apply to a terminal device with a low processing capability such as a mobile phone.

  The present invention has been made in view of the above problems, and when performing speech axis conversion by applying time axis companding to a speech signal, without overloading the apparatus that performs speech speed conversion, An object of the present invention is to provide a technique for avoiding the occurrence of waveform discontinuity in the processed audio signal.

  In order to solve the above-described problem, the present invention identifies a rising portion of a sound represented by an audio signal to be processed and a frame corresponding to the falling portion of the sound from a temporal change in the volume of the sound, A discriminating means for discriminating that a predetermined number of frames before and after the frame are speech speed conversion prohibiting frames for which speech speed conversion is prohibited, and for a frame determined by the discriminating means for being a speech speed conversion prohibiting frame A speech speed converting device comprising: a speech speed converting means for outputting the other frames while performing waveform insertion or waveform deletion so as to achieve the designated speech speed. provide.

  In order to solve the above-described problem, the present invention provides a computer apparatus with a rising portion of a sound represented by an audio signal to be processed and a frame corresponding to the falling portion of the sound from a time change of the sound volume. A first step of identifying the frame and a predetermined number of frames before and after the frame as a speech speed conversion prohibited frame for which speech speed conversion is prohibited; The frame determined in the first step is output as it is, while the other steps are output by performing waveform insertion or waveform deletion so as to achieve the designated speech speed. Provided is a program characterized in that

  According to the present invention, when the speech signal is subjected to time axis companding to convert the speech speed, the waveform of the processed speech signal is not discontinuous without imposing an excessive processing load on the device that performs the speech speed conversion. It is possible to avoid the occurrence of the occurrence of the problem.

Hereinafter, the best mode for carrying out the present invention will be described with reference to the drawings.
(A: Configuration)
FIG. 1 is a diagram illustrating an example of a hardware configuration of a speech speed conversion apparatus 10 according to an embodiment of the present invention.
A sound source (not shown) that outputs an analog audio signal S-in (t) is connected to the input terminal CH-in in FIG. 1, and an audio signal output from this sound source is input.
Although the detailed illustration is omitted in FIG. 1, the audio signal S-in (t) input to the input terminal CH-in is converted into a digital signal by an A / D conversion circuit (not shown) for a predetermined time. A long frame (8 milliseconds in this embodiment) is cut out in units of frames and delivered to the delay processing circuit 110 and the volume change rate calculation unit 120. Hereinafter, an audio signal converted into digital data and cut out in units of frames is also expressed as “S-in (t)”.
In the following, it is assumed that an audio signal S-in (t) having the signal waveform shown in FIG. 2A is input to the input terminal CH-in. In this embodiment, the case where the audio signal input to the input terminal CH-in is an analog signal will be described. However, it is a matter of course that the audio signal may be a digital signal. In this case, the A / D conversion circuit Needless to say, there is no need to provide a.

  The delay processing circuit 110 in FIG. 1 delays the input audio signal S-in (t) by a delay time Δt of about several tens of milliseconds and outputs the delayed signal to the speech speed conversion unit 140. Note that the reason why the delay process is performed on the audio signal S-in (t) is that the voice speed conversion is performed retroactively by the delay time Δt in order to make the user feel as if the voice speed conversion process is performed in substantially real time. It is for executing. Here, since the delay time Δt is about several tens of milliseconds, there is almost no influence on hearing.

The volume change rate calculation unit 120 in FIG. 1 sequentially receives the audio signal S-in (t) digitally converted as described above in units of frames, and represents a value representing the degree of temporal change in volume of the audio represented by the audio signal. ΔP (t) is calculated for each frame and delivered to the determination unit 130.
More specifically, the volume change rate calculation unit 120 specifies the volume of the audio signal for each frame by obtaining the envelope of the audio signal S-in (t) sequentially received in units of frames (FIG. 2 ( b)), and the first derivative (FIG. 2D) of the common logarithmic value of the sound volume (FIG. 2C) is calculated for each frame as ΔP (t) and delivered to the determination unit 130.
Note that in order to accurately specify the rising portion where the volume suddenly increases and the falling portion where the volume rapidly decreases in the sound represented by the audio signal S-in (t), the volume of the sound is expressed as ΔP (t). The first derivative of the common logarithm value (FIG. 2 (c)) is used. However, for example, when the rising portion and the falling portion can be specified using the first derivative of the sound volume, it is needless to say that the first derivative of the sound volume may be used as the ΔP (t).

Now, the volume change rate calculation unit 120 according to the present embodiment calculates ΔP (t) for the t-th frame according to the following equation (1).
ΔP (t) = {ln (10) × (P (t) −P (t−1))} / ln (P (t) (1)
In Equation (1), ln () means a natural logarithm, P (t) is a value obtained by normalizing the volume of the t-th frame with a predetermined reference volume, and P (t−1) Is a value obtained by normalizing the volume of the previous frame with a predetermined reference volume.
In the present embodiment, a case has been described in which the first-order differential of the sound volume is used as the value ΔP (t) indicating the degree of temporal change in volume of each frame, and the value is calculated according to the equation (1). However, other methods (for example, a method of obtaining the above ΔP (t) by referring to a table in which the values of the first derivative of the common logarithm with respect to the true values are stored in association with a plurality of different true values) Of course it is good.

Although not shown in detail in FIG. 1, the determination unit 130 includes a buffer having a storage capacity capable of storing a number of frames corresponding to the delay time Δt, and a comparator.
The determination unit 130 receives ΔP (t) delivered from the volume change rate calculation unit 120 and stores the ΔP (t) in the buffer in the order of reception. On the other hand, the change level of the volume at the past time by Δt from the ΔP (t). Is read from the buffer, and it is determined whether or not the magnitude of ΔP (t−Δt) exceeds a predetermined threshold th, and the value corresponding to the determination result is determined. A control signal CS is output.
More specifically, the determination unit 130 determines that a frame in which ΔP (t) has a value greater than the threshold th (a portion partitioned by a broken line in FIG. 2D) and a predetermined number of frames before and after the frame. (The portion indicated by hatching in FIG. 2D) is determined to be a frame for which speech speed conversion is prohibited (hereinafter referred to as “speech speed conversion prohibition frame”), and a signal value is given as a control signal CS indicating that. A control signal that is “1” (hereinafter referred to as a speech speed conversion prohibition signal) is output to the speech speed conversion unit 140. In addition, the determination unit 130 outputs a control signal CS (hereinafter referred to as a speech speed conversion permission signal) whose signal value is “0” to the speech speed conversion unit 140 for other frames. As a result, which of the plurality of frames constituting the audio signal S-in (t) is the speech rate conversion prohibited frame is transmitted to the speech rate conversion unit 140.

  The speech rate conversion unit 140 is, for example, a DSP, and corresponds to the speech rate specified by the user via an operation unit (not shown) according to the speech rate conversion algorithm (that is, PICOLA) disclosed in Non-Patent Document 1. A process for inserting or deleting a waveform corresponding to the number of cycles is applied to S-in (t-Δt), and an audio signal S-out (t-Δt) as a result of the processing is analogized by a D / A conversion circuit (not shown). The signal is converted into a signal and output to the outside via the output terminal CH-out. For example, a sound emitting device (not shown) such as a speaker is connected to the output terminal CH-out. From the sound emitting device, the sound signal subjected to the speech speed conversion processing by the speech speed converting device 10 is supplied. The corresponding sound is emitted.

However, the speech speed conversion unit 140 according to the present embodiment, when the control signal CS delivered from the determination unit 130 for the processing target frame is a speech speed conversion prohibition signal (that is, the processing target frame is the speech speed conversion prohibition). In the case of a frame), S-in (t-Δt) is directly output as S-out (t-Δt ′), which is different from the technique disclosed in Non-Patent Document 1. Note that the delay amount of the output signal S-out is “Δt ′”, which means that the delay amount can be different from Δt due to the speech speed conversion up to that point.
For this reason, in the speech rate conversion apparatus 10 according to the present embodiment, waveform insertion or waveform deletion is not performed in a portion where the volume is greatly changed (for example, a rising portion or falling portion of sound). The problem pointed out with respect to Patent Document 1 does not occur. Note that the consonant part of the plosive generally has a large degree of volume change, and the speech rate conversion prohibition signal is output by the determination unit 130 for such a part. The occurrence of problems such as multiple appearances is extremely low. Further, although the volume change is not always large for the consonant part of the plosive sound, even if the waveform discontinuity occurs in the part where the volume change is small, there is almost no effect on hearing.

As described above, according to the speech speed converting apparatus 10 according to the present embodiment, when speech speed conversion is performed by applying time axis companding to a speech signal, the waveform of the speech signal is reduced in a portion where the volume change is large. Continuation is avoided. In addition, in the speech speed conversion apparatus 10 according to the present embodiment, a part to be excluded from waveform insertion and waveform deletion targets is specified based on the rate of change in volume, and complicated processing such as voice recognition is not performed. Therefore, an excessive load is not applied to the speech speed conversion apparatus 10.
Thus, according to the present embodiment, when the speech signal is subjected to time axis companding to convert the speech speed, the processed speech signal is not subjected to an excessive processing load on the device that performs the speech speed conversion. It is possible to avoid the occurrence of discontinuity in the waveform.

(B: Deformation)
Although one embodiment of the present invention has been described above, it is needless to say that the embodiment may be modified as described below.
(1) In the above-described embodiment, a case has been described in which speech speed conversion is prohibited for a frame in which the magnitude of ΔP (t) exceeds a predetermined threshold and a predetermined number of frames before and after the frame. However, the length of the section in which speech speed conversion is prohibited is changed between the rising part of the sound (the part where ΔP (t) is positive) and the falling part of the sound (for example, speaking speed conversion is performed for the falling part of the sound). It is also possible to lengthen the prohibited section.

(2) In the above-described embodiment, the calculation means (volume change rate calculation unit 120) that calculates a value indicating the time change rate of the logarithmic value of the sound volume represented by the audio signal to be processed for each frame, Among a plurality of constituting frames, it is determined that speech rate conversion is prohibited for a frame whose magnitude calculated by the calculation means exceeds a predetermined threshold and a predetermined number of frames following the frame. The discriminating means (discriminating unit 130) outputs the frame as it is determined that the speech speed conversion is prohibited, while performing waveform insertion or waveform deletion so as to achieve the designated speech speed for the other frames. Each of the speech speed converting means (speech speed converting unit 140) that outputs the data is realized by hardware modules, and the hardware modules are combined to provide the present invention. Been described to configure the speed converter, it is of course may be to realize these respective means in the software module.

  Specifically, a program that causes a CPU (Central Processing Unit) to function as the volume change rate calculation means, the determination means, and the speech speed conversion means (for example, a program that causes the CPU to execute the speech speed conversion processing shown in FIG. 3). It may be installed in a general computer device such as a personal computer so that the computer device functions as the speech rate conversion device according to the present invention.

It is a block diagram which shows the structural example of the speech speed converter 10 which concerns on one Embodiment of this invention. It is a figure which shows the process of the speech speed conversion process which concerns on this embodiment. It is a flowchart which shows the flow of the speech speed conversion process which concerns on a modification (1).

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 ... Speech speed conversion apparatus, CH-in ... Input terminal, 110 ... Delay circuit, 120 ... Volume change rate calculation part, 130 ... Discrimination part, 140 ... Speech speed conversion part, CH-out ... Output terminal.

Claims (4)

The frame corresponding to the rising portion of the sound represented by the audio signal to be processed and the falling portion of the sound is identified from the temporal change in the volume of the sound, and the frame and a predetermined number of frames before and after the frame are talked about. A discriminating means for discriminating that the frame is a speech speed conversion prohibition frame prohibited from speed conversion;
While the frame determined by the determination unit as being a speech rate conversion prohibition frame is output as it is, the other frames are output by performing waveform insertion or waveform deletion so that the specified speech rate is obtained. Conversion means;
A speech rate conversion device comprising: A calculation unit that calculates a value indicating a time change rate of a logarithmic value of a volume of a voice represented by the voice signal for each frame;
The discrimination means includes
Of the plurality of frames constituting the audio signal, the speech rate conversion prohibition frame for a frame whose magnitude calculated by the calculation means exceeds a predetermined threshold and a predetermined number of frames before and after the frame. It is discriminate | determined that it is. The speech-speed converter of Claim 1 characterized by the above-mentioned. A calculation means for calculating a value obtained by dividing a time change of a volume of a voice represented by the voice signal by the volume, for each frame;
The discrimination means includes
Of the plurality of frames constituting the audio signal, the speech rate conversion prohibition frame for a frame whose magnitude calculated by the calculation means exceeds a predetermined threshold and a predetermined number of frames before and after the frame. It is discriminate | determined that it is. The speech-speed converter of Claim 1 characterized by the above-mentioned. Computer equipment,
The frame corresponding to the rising portion of the sound represented by the audio signal to be processed and the falling portion of the sound is identified from the temporal change in the volume of the sound, and the frame and a predetermined number of frames before and after the frame are talked about. A first step of determining that the frame is a speech speed conversion prohibited frame in which speed conversion is prohibited;
The frame determined in the first step is output as it is if it is a speech rate conversion prohibition frame, while the other frames are output after waveform insertion or waveform deletion so that the designated speech rate is obtained. A second step of:
A program characterized in that is executed.

Images