JP2001109499A - Speech speed conversion device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent a slowly spoken sound signal from being converted into a more slowly spoken sound signal resulting from applying a speech speed conversion processing to the slowly spoken sound signal in a speech speed conversion device. SOLUTION: A pitch period Tn-1 stored in a pitch period storage part 2 is compared with a pitch period Tn which is newly extracted in a pitch period extraction part 1 by a pitch period comparison part 3. The value of a counter 4 is increased in accordance with the comparison result. A comparison part 6 compares a value M obtained by multiplying the pitch period Tn by the value C of the counter 4 with a threshold S which is set in a threshold setting part 5. A speech speed conversion part 7 applies a speech speed conversion processing to an input sound signal at a prescribed mode based on the comparison result.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech rate converter for changing the speech rate of an audio signal. For the hearing impaired and the elderly, hearing aids and devices such as telephones equipped with the hearing aids, which convert broadcast audio signals into slow, easy-to-hear voices, and English voice spoken at native speed slowly The present invention relates to a speech speed conversion device that can be used in various devices, such as an English language learner that converts speech into an easy-to-hear speech.

The speech speed conversion means that the time axis of an audio signal is compressed to make its reproduction speed faster than the original speed, or conversely, the time axis of the audio signal is expanded to make its reproduction speed lower. Or slower than your speed.

[0003]

2. Description of the Related Art Conventionally, as disclosed in, for example, Japanese Patent Application Laid-Open No. 7-192392, a compression / expansion process is performed on an input audio signal depending on whether the input audio signal is a voice section or a silent section. Alternatively, a speech speed conversion device configured to perform a deletion process is known.

[0004]

However, in the above-mentioned conventional speech speed conversion device, the speech speed is uniformly reduced regardless of the speech speed of the input voice signal. However, there is a problem that the speech speed conversion device further slows down the operation, and conversely, even if the speech speed is originally fast and fast, the speech speed conversion device further increases the speed.

[0005]

In order to solve the above-mentioned problems, in a speech speed conversion device according to the present invention, a pitch cycle detecting means for detecting a pitch cycle from a voice signal and a pitch cycle extracted by the pitch cycle detecting means are provided. Counting means for counting the number of repetitions of a predetermined pitch cycle based on the comparison, and a comparison determination means for comparing a product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means with a predetermined threshold value And a speech speed conversion means for performing speech speed conversion based on the result of the judgment by the comparison judgment means.

Further, in the speech speed conversion device of the present invention, a pitch cycle detecting means for detecting a pitch cycle from a voice signal, and counting the number of repetitions of a predetermined pitch cycle based on the pitch cycle extracted by the pitch cycle detecting means. Counting means,
A comparison determining unit that compares a product of the pitch period extracted by the pitch period detecting unit and the number of repetitions counted by the counting unit with a predetermined threshold; a threshold changing unit that changes the predetermined threshold; Speech speed conversion means for performing speech speed conversion based on the determination result of the means.

Further, the speech speed conversion means is provided only when the product of the pitch cycle extracted by the pitch cycle detection means in the comparison determination means and the number of repetitions counted by the counting means does not exceed a predetermined threshold value. It is characterized in that the speech signal is converted into a slow speech signal by speaking speed.

Further, the speech speed conversion means determines that the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means does not exceed a predetermined threshold value. The voice signal is converted into a slow voice signal and the speech speed is converted, and the product of the pitch cycle extracted by the pitch cycle detecting means and the number of repetitions counted by the counting means in the comparing and determining means exceeds a predetermined threshold value. When the determination is made, the speech speed conversion is not performed.

If the product of the pitch cycle extracted by the pitch cycle detecting means and the number of repetitions counted by the counting means is determined to be within a predetermined threshold value, the speech speed converting means determines the speech rate. If the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means is determined to be greater than a predetermined threshold, the speech rate magnification is decreased. The speech rate conversion is performed (however, the speech rate magnification = time length of input voice signal / time length of output voice signal).

The speech speed conversion apparatus according to the present invention comprises: a pitch cycle detecting means for detecting a pitch cycle from a voice signal; and counting the number of repetitions of a predetermined pitch cycle based on the pitch cycle extracted by the pitch cycle detecting means. Counting means,
Comparison determination means for comparing the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means with a predetermined first threshold value and a predetermined second threshold value; Threshold value <second threshold value), and a speech speed conversion unit that performs speech speed conversion based on the determination result of the comparison determination unit.

Further, the speech speed conversion apparatus according to the present invention comprises a pitch cycle detecting means for detecting a pitch cycle from a voice signal, and counting the number of repetitions of a predetermined pitch cycle based on the pitch cycle extracted by the pitch cycle detecting means. Counting means, and comparison determining means for comparing a product of the pitch cycle extracted by the pitch cycle detecting means and the number of repetitions counted by the counting means with a predetermined first threshold value and a predetermined second threshold value ( However, the threshold value changing means for changing the first threshold value <the second threshold value), the predetermined first threshold value or the predetermined second threshold value, and the speech speed conversion is performed based on the determination result of the comparison determination means. And a speech speed conversion unit.

[0012] The speech speed conversion means may include a product of the pitch cycle extracted by the pitch cycle detection means in the comparison determination means and the number of repetitions counted by the counting means being a predetermined first threshold value and a predetermined first threshold value. If it is determined that the difference is between the threshold values of 2, the speech speed conversion is not performed.

[0013] The speech speed conversion means may determine that the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means is smaller than a predetermined first threshold value in the comparison determination means. When it is determined, the speech speed conversion is performed with the speech speed magnification reduced (however, the speech speed magnification = time length of input voice signal / time length of output voice signal).

Further, the speech speed conversion means may determine that the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means in the comparison determination means is greater than a predetermined second threshold value. When the judgment is made, the speech speed conversion is performed by increasing the speech speed magnification (however, the speech speed magnification = time length of input voice signal / time length of output voice signal).

[0015] The speech speed conversion means may determine that the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means is smaller than a predetermined first threshold value in the comparison determination means. When it is determined, the speech speed conversion is performed with the speech speed ratio smaller than 1 (however, the speech speed ratio = time length of input audio signal / time length of output audio signal).

The speech speed converting means may be arranged such that the product of the pitch cycle extracted by the pitch cycle detecting means and the number of repetitions counted by the counting means is larger than a second predetermined threshold value by the comparing and judging means. When it is determined, the speech speed conversion is performed with the speech speed magnification greater than 1 (however, speech speed magnification = time length of input audio signal / time length of output audio signal).

Further, the speech speed conversion means changes the speech speed magnification in accordance with the empty capacity of the storage means for storing the speech signal whose speech speed has been converted (where speech speed magnification = input voice). Signal time length / output audio signal time length).

Further, the speech speed conversion means changes the speech speed magnification to be closer to 1 as the empty capacity of the storage means for storing the speech signal whose speech speed has been converted decreases. (However, speech speed magnification = time length of input audio signal / time length of output audio signal).

Further, the speech speed conversion means changes the speech speed magnification closer to a predetermined magnification in accordance with an increase in the empty capacity of the storage means for accumulating the speech signal whose speech speed has been converted. (However, speech speed magnification = time length of input audio signal / time length of output audio signal).

Further, the predetermined pitch cycle is the same pitch cycle, a double pitch cycle or a half pitch cycle, and a pitch cycle approximate to these.

[0021]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of a speech speed converter according to the present invention.

FIG. 1 is a schematic block diagram showing the configuration of a speech speed conversion device according to the present invention. In FIG. 1, reference numeral 1 denotes a pitch cycle extracting unit which receives a voice signal converted into a digital signal by an A / D converter (not shown) and extracts a pitch cycle from the input voice signal. For example, an autocorrelation is used as an extraction method.

The pitch period extraction method using autocorrelation includes:
Assuming that the signal is time-limited, there is a method using short-time autocorrelation in which a signal exists only within the section of the time length Ts and the autocorrelation is always set to zero outside the section of the time length Ts. . This is a digital signal processing of voice issued by Corona (above)-LRRabiner & R.W.Schafer
As described in the book, Kuki Suzuki-p152-p152, if the voice waveform is represented by digital voice data x (n), the short-time autocorrelation value Rn (k) by the above-described method is as follows. Become like

[0024]

(Equation 1)

Here, Ts is a time section in which a voice signal is assumed to exist, k is a delay time for delaying a voice waveform when calculating a short-time autocorrelation value Rn (k), and T
There is a relationship of s≫k. When the value of k that maximizes the short-time autocorrelation value Rn (k) in Equation 1 is obtained, the value is the pitch period.

Next, reference numeral 2 denotes a pitch period storage unit in which the pitch periods extracted by the pitch period extraction unit 1 are stored, and 3 denotes a pitch period stored in the pitch period storage unit 2 and the pitch period extraction unit 1. A pitch cycle comparing section for comparing the pitch cycle with the pitch cycle newly extracted in step 4;
Is a counter that is incremented in accordance with the comparison result in, and 5 is a threshold value setting unit in which a threshold value (details will be described later) is set in advance.

6 compares a value obtained by multiplying the pitch period extracted by the pitch period extracting unit 1 with the value of the counter 4 and a threshold value set by the threshold value setting unit 5; , A speech rate conversion section that performs speech rate conversion processing in a predetermined mode on the basis of the comparison result output from the comparison section 6 and outputs the speech signal, and 8 a speech rate conversion section. A mode selection unit that outputs a mode selection signal for selecting a mode (details will be described later).

FIG. 2 is a schematic block diagram showing the configuration of the speech speed conversion section 7. As shown in FIG.

In FIG. 1, reference numeral 11 denotes an audio time axis compression / expansion unit for compressing / expanding the time axis of an input audio signal. The compression / expansion method used here is, for example, overlap addition by pointer movement amount control. Law (Pointer Interv
al Control Overlap and Add: PICOLA) and TDH
Known methods such as the S (Time Domain Harmonic Scaling) method can be used, but are not limited to these methods. In short, any method can be used as long as the time axis of the audio signal can be compressed and expanded to change its reproduction speed.

Next, reference numeral 14 denotes the audio time axis compression / expansion unit 1
1, an audio encoder for encoding the audio signal subjected to the compression / expansion processing by the existing ADPCM processing, etc., 15 is a memory for storing the signal encoded by the audio encoder 14, and 16 is the memory 15 From existing ADPC
This is a speech decoding unit for decoding by M processing or the like. The audio signal decoded by the audio decoding unit 16 is converted into an analog audio signal by a D / A conversion circuit (not shown) and output.

Reference numeral 12 denotes a silent section detecting section for detecting a silent section in the input audio signal and transmitting the detection result to the audio time axis compression / expansion section. A speech speed control unit 17 for giving a magnification N (double speed) is an accumulation amount detection unit for detecting an accumulation amount j of the signal data in the memory 15.

Here, the speech speed magnification N is expressed by [speech speed magnification N] (double speed) = [time length of input voice signal] / [time length of output voice signal], and the time length of the input voice signal Is
The time length of the audio signal before compression / expansion input to the audio time axis compression / expansion unit 11, and the time length of the output audio signal refers to the compression time of the input audio signal decoded by the audio decoding unit 16. This is the length of time after elongation.

Next, various modes of the speech speed which are selected by the mode selection unit 8 in FIG. 1 and which are subjected to the speech speed conversion processing by the speech speed control unit 13 based on the selection will be described.
As the speech speed mode, as shown in Table 1 below, there is provided a [slow playback mode] in which the time axis of the input audio signal is extended to convert the speech speed into a slow speech signal. As shown in the table, the speech speed magnification N can be finely selected in 1 to 4 stages.

[0034]

[Table 1]

The value of the speech speed magnification N is further finely classified according to the accumulation amount j of the memory 15 given from the accumulation amount detection unit 17, but the accumulation amount j of the memory 15 and the speech speed magnification The relationship with N will be described later.

In each table, the speech speed magnification N is represented by [time length of input voice signal] / [time length of output voice signal] as described above. Indicates that the speech speed is lower than the normal speed, and the smaller this value is, the lower the speech speed is. Conversely, when the speech speed magnification N is greater than 1, it indicates that the speech speed is faster than the normal speed, and the larger the value, the faster the speech speed.

By the way, the selection of the slow reproduction mode is performed, for example, in the case of [50s], [60s], [70s], [8].
For example, four operating buttons labeled according to the age of the user may be selected.

Supplementally, "Evaluation experiments on speech rate converted speech for elderly people: (Atsushi Imai, Nobumasa Kiyoyama, Toru Toki, Eiichi Miyasaka, Hiroshi Ono), Proceedings of the Acoustical Society of Japan, March 1993. ) Describes the evaluation of the speech speed of the converted speech and the ease of hearing by age.According to this paper, the speech speed is 0.87 times faster in the 60s and 0 in the 70s. .8
It has been found that it is easy to hear at 0.73 × speed in the 7s to 0.73 × speed and 80s in the 80s, and it is known that as the age of the listener increases, the speaking speed at which the listener feels easy to hear also decreases.

In the speech speed conversion unit configured as described above, the speech speed control unit 13 includes the mode selection signal from the mode selection unit 8 shown in FIG. 1, the comparison result from the comparison unit, Based on information such as the storage amount j from the storage amount detection unit 17, the speech speed magnification N is determined and output to the audio time axis compression / expansion unit 11.

The voice time axis compression / expansion unit 11 performs a process of compressing or expanding the time axis of the audio signal based on the voice speed magnification N from the voice speed control unit 13. Then, the audio time axis compression / expansion unit 11 receives the detection result of the silent section of the audio signal by the silent section detection unit 12 and, based on this, appropriately deletes the silent section and compresses or compresses the time axis of the audio signal. Perform decompression processing.

Next, the operation of the first embodiment of the speech speed conversion device configured as described above will be described with reference to the flowchart of FIG.

First, the user operates the operation button (not shown) of the mode selection section 8 to select a mode (S1).
1). Here, as an example, it is assumed that [Slow playback mode 1] shown in Table 1 is selected. By the above operation, the mode selection signal is given from the mode selection unit 8 to the speech speed control unit 13 of the speech speed conversion unit 7.

Next, the storage amount detector 17 checks the storage amount in the memory 15 to determine the storage amount j, and gives the value of the storage amount j to the speech speed controller 13 (S12).

At this time, assuming that the accumulated amount of the encoded audio signal in the memory 15 is 0, 0.6 [times] is set as the initial value as the speech speed magnification N from Table 1 (see FIG. 1). S13).

The contents stored in the pitch period storage unit 2 (hereinafter, referred to as a pitch period Tn-1 as a variable name) are initialized (cleared).

Thus, the speech speed magnification N (=
0.6 [double speed]), the pitch period extracting unit 1
Extracts the pitch cycle of the input audio signal (hereinafter, referred to as pitch cycle Tn as a variable name) based on Equation 1 (S14). As an example, for example, pitch period Tn = 6
It is assumed that 0 [sample] is obtained. The “sample” means the number of audio signals sampled according to a desired sampling frequency when the audio signal is a digital signal.

The pitch cycle Tn extracted by the pitch cycle extracting section 1 is compared with the pitch cycle Tn-1 stored in the pitch cycle storing section 2 by the pitch cycle comparing section 3 (S1).
5).

However, since the contents stored in the pitch period storage unit 2 have been cleared as described above, step S2
The process proceeds to 1 to initialize (clear) the count value C of the counter 4 to 0, and to set the [slow reproduction mode] in the next step S19 (already the step S11).
Since [Slow playback mode] has been selected in [], the mode is not substantially changed, and [Slow playback mode] is selected.
Will be maintained).

The value of the pitch period Tn (= 60 [samples]) is stored in the pitch period storage unit 2 and becomes a new value of the pitch period Tn-1 (S22). The speech speed magnification N = 0.6 [double speed] is given to the audio time axis compression / expansion unit 11, and the time axis compression / expansion unit 11 receives the input signal so that the speech speed becomes 0.6 [double speed]. The time axis of is extended. The audio signal whose time axis has been expanded by the time axis compression / expansion unit 11 is encoded by an audio encoding unit 14, and once stored in a memory 15.
The audio signal is decoded by the audio decoding unit 16 to become an output audio signal, and the process returns to step S12 via step S23.

Then, the storage amount j of the memory is checked again (S12), and the speech speed magnification N based on the storage amount j is set (S13).

Then, it is assumed that the pitch period Tn is further extracted (S14), and for example, a pitch period Tn = 61 [samples] is obtained.

The extracted pitch cycle Tn (= 61) and the pitch cycle Tn-1 stored in the pitch cycle storage unit 2
(= 60 [samples)) is compared by the pitch period comparison unit 3 (S15).

Here, the newly extracted pitch period Tn
(= 61 [sample]) and the pitch cycle Tn-1 (= 60 [sample]) stored in the pitch cycle storage unit 2
The condition Tn ≒ Tn-1 (that is, the newly extracted pitch period Tn is substantially equal to the previously extracted pitch period Tn-1)
Is satisfied, the count value C of the counter 4 is incremented by one (count value C = 1) (S1).
6).

Then, the pitch period Tn (= 61 [sample]) extracted by the pitch period extracting unit 1 and the count value C (= 1) of the counter 4 are integrated to obtain an integrated value M (= 61). This is given to the comparison unit 6 (S17).

The comparing section 6 includes the integrated value M and the threshold value setting section 5
Is compared with the threshold value S (= 1500) set in step (S18).

As described above, since the integrated value M is equal to 61 and does not exceed the threshold value S, the speech speed control unit 13 maintains the speech speed mode in the "slow reproduction mode", and j is given to the speech time axis compression / expansion unit 11, and the time axis compression / expansion unit 11 receives the speech rate N and sets the time of the input audio signal to the given speech rate magnification N. Extend the axis.

As described above, the time axis compression / expansion unit 11
The audio signal whose time axis has been expanded in step (1) is encoded by the audio encoding unit 14, stored at one end in the memory 15, and then decoded by the audio decoding unit 16 to become an output audio signal.

Then, the value of the pitch period Tn is newly stored in the pitch period storage section 2 (S22), and the process returns to step S12 via step S23.

Step S12 as described above →→
In the loop processing that returns to step 12 after step S23, if the input audio signal is a speech signal when the speech is slowly spoken, the loop processing is repeated, and in step S18, the integrated value M exceeds the threshold value S. .

FIG. 3 (c) shows a voice signal when the user speaks high and slowly, and shows a waveform in which the same pitch period Tn is repeated 10 times. FIG. And the same pitch period Tn is 4
This is a waveform repeated twice. Speech signals when spoken slowly are sounded one by one (""
The duration of “a”, “i”, “u”,... Is long, and in the case of such a waveform, the increment of the counter 4 is repeated until the integrated value M reaches the threshold value S (= 150).
0), and goes through step S18 to step S20.
And the normal playback mode (speech rate magnification N = 1.0) is reached, and speech rate conversion is not substantially performed.

FIG. 3A is a waveform of a high-pitched and fast-talking voice signal in which the same pitch period Tn is repeated four times. FIG. 3B is a low-pitched and fast-talking voice signal. The signal is a waveform in which the same pitch cycle Tn is repeated twice. In the case where the input audio signal is a fast-speech sound signal (speech signal when the user speaks fast), as shown in FIGS.
Therefore, even if the loop processing is repeated, the pitch period extraction processing must be performed before the integrated value M exceeds the threshold value S, even if the loop processing is repeated. In step S15, the pitch cycle Tn newly extracted (eg, the pitch cycle of the sound “I”) and the pitch cycle Tn−1 (eg, “1”) extracted immediately before
The pitch cycle of the sound "A") is different.

Therefore, the process proceeds to step 21 via step S15, where the count value C of the counter 4 is cleared. Further, the process proceeds to step S19, where the voice signal of the fast voice continues regardless of whether the voice is high or low. Means that the [slow playback mode] is maintained.

As described above, the present invention is characterized in that the voice signal of the fast-talking voice has a short duration of each sound, and the voice signal when speaking slowly has a long duration of each sound. By focusing on the product of the number of repetitions of the waveform of the same pitch cycle and the pitch cycle and comparing it with a predetermined threshold value, without being affected by the high or low voice, the voice signal spoken quickly or spoken slowly It is determined whether the signal is an audio signal, and a speech speed conversion process is performed so that only the fast-speech audio signal becomes a slow audio signal. This is used to change the speech rate magnification of the speech rate conversion.

By the way, the comparison condition in the pitch period comparison unit 3 in the step S15 is Tn ≒ Tn-1 and the like. The reason will be described below.

A human voice is different from a stable signal such as a sine wave generated in an electronic circuit, and a pitch or the like often fluctuates. Therefore, even if the pitch period is sequentially obtained for the same sound, the value may slightly differ depending on the obtained time. In such a case, the comparison condition is set as described above in order to prevent an erroneous determination that the sound has changed to a different sound even though the same sound continues.

The allowable range in which the newly extracted pitch period Tn and the immediately preceding pitch period Tn-1 are determined to be substantially the same is also affected by the sampling frequency. The higher the sampling frequency, the wider the allowable range is set. There is a need to. In this embodiment, the sampling frequency fs = 12.8 KHz
And the allowable range was set to 3 or less.

When the pitch period is extracted by the pitch period extracting unit 1, when a new pitch period is extracted due to the influence of fluctuations of the pitch and the like as described above, the pitch period of the immediately preceding pitch period is extracted. A value that is approximately twice or approximately half may be extracted. For this reason, as described in the flowchart, it is configured that the same pitch period is also determined when Tn ≒ 2Tn-1 or Tn ≒ 1 / 2Tn-1.

Next, with respect to the processing in step S12, the relationship between the storage amount j of the memory 15 obtained by the storage amount detection unit 17 and the speech speed magnification N will be described.

As the speech speed conversion process in the speech speed conversion unit 7 continues, the vacant capacity of the memory 15 decreases as the encoded voice signals are accumulated in the memory 15. Memory 1
5 is configured to hold a coded digital audio signal for a fixed amount. However, if there are few silent sections to be deleted in the input audio signal, the audio signal can no longer be stored in the memory 15 and is not stored. There is a possibility that an inconvenience such as a missing audio signal is lost. In order to avoid such inconvenience, the storage amount of the memory 15 is checked, and the speech speed magnification N is corrected as the remaining amount of the memory 15 decreases.

In the above example, the initial value of the speech speed magnification N is set to 0.6 [double speed]. However, as shown in Table 1, as the storage amount j of the memory 15 increases, the speech speed magnification N Shift the value of to the right. Specifically, while the loop processing is repeated, in step 12, when the accumulated amount j becomes a value in the range of 20 ≦ k <40 [%], the speech speed magnification N is increased from 0.6 [double speed]. Change to 0.7 [double speed]. Thereafter, the value is changed to the value shown in the table according to the accumulated amount j, and the speech speed ratio N = 1.0 before the memory 15 overflows.
[Double speed] (normal playback without compression / expansion on the time axis = same as playback in [normal mode]).

On the other hand, when the input voice signal includes many silent sections that can be deleted, the storage amount j of the memory 15 gradually decreases, and in Table 1, the value of the speech speed magnification N is changed in the opposite manner. When the value shifts to the left, and the value of the speech speed magnification N reaches the initial value, the shift of the value of the speech speed magnification N stops.

When a stop instruction is given by the user, the apparatus stops (S23).

In summary, in this embodiment, when the integrated value M is larger than the threshold value S, the [normal reproduction mode] is automatically selected, and when the integrated value M is less than the threshold value S, the [slow reproduction mode] is set. It is automatically selected.

Next, a second embodiment of the present invention will be described with reference to the flowchart of FIG. The basic configuration of the device is the same as that shown in FIGS.
The detailed description is omitted, and only the operation is described.

First, the user operates the operation button (not shown) of the mode selection section 8 to select a mode as in the above embodiment (S31). As an example, it is assumed that [Slow playback mode 1] shown in Table 1 is selected.
As a result, the mode selection signal is given from the mode selection unit 8 to the speech speed control unit 13 of the speech speed conversion unit 7, and the speech speed control unit 1
3 is set to [slow reproduction mode 1] based on the given mode selection signal.

Next, the storage amount detector 17 checks the storage amount in the memory 15 to obtain the storage amount j, and gives the value of the storage amount j to the speech speed controller 13 (S32).

At this point, assuming that the accumulated amount of the coded audio signal in the memory 15 is 0, from the above Table 1, the speech speed magnification N is set to 0.6 [double speed] as an initial value. (S33).

The contents stored in the pitch period storage unit 2 (hereinafter, referred to as a pitch period Tn-1 as a variable name) are initialized (cleared).

Thus, the speech speed magnification N (=
0.6 [double speed]), the pitch period extracting unit 1
Extracts the pitch period of the input audio signal (hereinafter referred to as pitch period Tn as a variable name) based on the above equation (S34). As an example, for example, pitch period Tn = 6
It is assumed that 0 [sample] is obtained.

The pitch cycle Tn extracted by the pitch cycle extracting section 1 is compared with the pitch cycle Tn-1 stored in the pitch cycle storing section 2 by the pitch cycle comparing section 3 (S3).
5).

However, since the contents stored in the pitch period storage unit 2 have been cleared as described above, step S4
The process proceeds to 1 to initialize (clear) the count value C of the counter 4 to 0.

The value of the pitch cycle Tn (= 60 [sample]) is stored in the pitch cycle storage unit 2 and becomes a new value of the pitch cycle Tn-1 (S42). The speech speed magnification N = 0.6 [double speed] is given to the audio time axis compression / expansion unit 11, and the time axis compression / expansion unit 11 receives the input signal so that the speech speed becomes 0.6 [double speed]. The time axis of is extended. The audio signal whose time axis has been expanded by the time axis compression / expansion unit 11 is encoded by an audio encoding unit 14, and once stored in a memory 15.
The audio signal is decoded by the audio decoding unit 16 to become an output audio signal, and the process returns to step S32 via step S43.

Then, the storage amount j of the memory is checked again (S32), and the speech speed magnification N based on the storage amount j is set again (S33).

Then, it is assumed that the pitch period Tn is extracted again (S34), and for example, the pitch period Tn = 61 [samples] is obtained. This extracted pitch cycle Tn
(= 61 [sample]) and the pitch period Tn-1 (= 60 [sample]) stored in the pitch period storage unit 2 are compared by the pitch period comparison unit 3 (S35).

Here, the newly extracted pitch period Tn
(= 61 [sample]) and the pitch cycle Tn-1 (= 60 [sample]) stored in the pitch cycle storage unit 2
If the condition of Tn ≒ Tn-1 is satisfied, the count value C of the counter 4 is incremented by one (S36).

Then, the pitch period Tn (= 61 [sample]) extracted by the pitch period extracting unit 1 and the count value C (= 1) of the counter 4 are integrated to obtain an integrated value M (= 61) (S37). Are given to the comparison unit 6.

The comparing section 6 compares the integrated value M with the threshold value setting section 5
Is compared with the threshold value S (= 1500) set in step (S38). As described above, since the integrated value M is 61 and does not exceed the threshold value S, the speech speed control unit 13 maintains the speech speed mode in the [slow playback mode], and according to the memory accumulation amount j. Is given to the speech time axis compression / expansion unit 11, and the time axis compression / expansion unit 11 receives the speech rate magnification N and adjusts the time axis of the input audio signal so that the time axis is doubled according to the speech rate magnification N. Elongate.

As described above, the time axis compression / expansion unit 11
The audio signal whose time axis has been expanded in step (1) is encoded by the audio encoding unit 14, stored at one end in the memory 15, and then decoded by the audio decoding unit 16 to become an output audio signal.

Then, the value of the pitch cycle Tn is newly stored in the pitch cycle storage section 2 (S42), and the process returns to step S32 via step S43.

In the above-described loop processing of step S32 →... →→ step S43 and returning to step S32, if the input audio signal is a speech signal in the case of a slow speech, the loop processing is repeated, and then step S32
At 38, the integrated value M exceeds the threshold value S set by the threshold value setting unit 5, and the process proceeds to step S40. In step 40, the speech speed magnification N in Table 1 is shifted right by one from the current value. As a result, the expansion rate of the time axis of the input audio signal becomes slightly smaller than before, and the speech speed becomes slightly faster than before. That is, the value (the speech speed magnification N) in Table 1 is shifted from the current value to the next value and set.

Conversely, when the input audio signal is a fast-speech audio signal, since the integrated value M does not exceed the threshold value S set by the threshold value setting section 5, the process proceeds from step S38 to step 39. In step 39, the speech speed magnification N in Table 1 is shifted left by one from the current value. As a result, the expansion rate of the time axis of the input audio signal becomes slightly larger than before, and the speech speed becomes slightly lower than before.

It is to be noted that the speech speed magnification N is also changed depending on the storage amount k of the memory as described above.

When the user gives a stop instruction, the apparatus stops (S43).

In summary, in this embodiment, when the integrated value M is smaller than the threshold value S, [slow reproduction mode]
Is automatically selected to slightly increase the expansion rate in the case of, and when the integrated value M is larger than the threshold value S, the expansion rate in the [slow playback mode] is automatically selected to be slightly reduced.

Next, a third embodiment of the present invention will be described with reference to the flowchart of FIG. Note that the basic configuration of the apparatus is the same as that shown in FIGS. 1 and 2 and detailed description thereof is omitted. However, in the present embodiment, the threshold setting unit 5 uses the first threshold S1 and the first threshold S1. And a second threshold value S2 set to a large value.

In FIG. 6, first, the user operates the operation button (not shown) of the mode selection section 8 to select a mode as in the above embodiment (S51). As an example, it is assumed that [Slow playback mode 1] shown in Table 1 is selected. As a result, the mode selection signal is given from the mode selection unit 8 to the speech speed control unit 13 of the speech speed conversion unit 7, and the speech speed control unit 13 sets [slow playback mode 1] based on the given mode selection signal. Is done.

Next, the storage amount detector 17 checks the storage amount in the memory 15 to determine the storage amount j, and gives the value of the storage amount j to the speech speed controller 13 (S52).

At this point, assuming that the accumulated amount of the encoded audio signal in the memory 15 is 0, from the above Table 1, the speech speed magnification N is set to 0.6 [double speed] as the initial value ( S53) The speech speed conversion processing is performed at the speech speed magnification N = 0.6 [double speed]. The audio signal whose time axis has been expanded by the time axis compression / expansion unit 11 is encoded by an audio encoding unit 14, stored in a memory 15 at one end, and then decoded by an audio decoding unit 16 to output an output audio signal. Becomes

The contents stored in the pitch period storage unit 2 (hereinafter, referred to as a pitch period Tn-1 as a variable name) are initialized (cleared).

Thus, the speech speed ratio N (=
0.6 [double speed]), the pitch period extracting unit 1
Extracts the pitch period of the input audio signal (hereinafter, referred to as pitch period Tn as a variable name) based on Equation 1 (S54). As an example, for example, pitch period Tn = 6
It is assumed that 0 [sample] is obtained.

The pitch period Tn extracted by the pitch period extraction unit 1 is compared with the pitch period Tn-1 stored in the pitch period storage unit 2 by the pitch period comparison unit 3 (S5).
5).

However, since the contents stored in the pitch period storage unit 2 have been cleared as described above, step S6 is executed.
2 and the count value C of the counter 4 is initialized (cleared) to 0, and the value of the pitch period Tn (=
60 [sample]) is stored in the pitch period storage unit 2 and becomes a new value of the pitch period Tn-1 (S63), and the process returns to step S52 via step S63.

Then, the storage amount j of the memory is checked again (S52), and the speech speed magnification N based on the storage amount j is set again (S53).

Then, it is assumed that the pitch period Tn is extracted again (S54), and for example, the pitch period Tn = 61 [samples] is obtained. This extracted pitch cycle Tn
(= 61 [sample]) and the pitch period Tn-1 (= 60 [sample]) stored in the pitch period storage unit 2 are compared by the pitch period comparison unit 3 (S55).

Here, the newly extracted pitch period Tn
(= 61 [sample]) and the pitch cycle Tn-1 (= 60 [sample]) stored in the pitch cycle storage unit 2
If the condition of Tn ≒ Tn-1 is satisfied, the count value C of the counter 4 is incremented by one (S56).

The pitch period Tn (= 61 [sample]) extracted by the pitch period extracting section 1 and the counter 4
And the integrated value M (= 61)
(S57), and this is given to the comparison unit 6.

The comparing section 6 compares the integrated value M with the threshold value setting section 5
Are compared with the first threshold value S1 (for example, 1000) and the second threshold value S2 (for example, 2000) (S58). As described above, since the integrated value M = 61, it is smaller than the first threshold value S1, and the speech speed control unit 13 changes the input signal expansion rate to be slightly higher (S59). That is, in Table 1, the current value (the speech speed magnification N) is shifted to the left by one and set.

As in the above embodiments, the audio signal whose time axis has been expanded by the time axis compression / expansion unit 11 is encoded by the audio encoding unit 14 and once stored in the memory 15. The audio signal is decoded by the audio decoding unit 16 and becomes an output audio signal.

Then, the value of the pitch cycle Tn is newly stored in the pitch cycle storage unit 2 (S63), and the process returns to step S52 via step S63.

In the loop processing that returns to step 52 via step S52 → step S64 as described above, the loop processing is repeated, and the integrated value M becomes between the first threshold value S1 and the second threshold value S2. If so, step S6
However, in this case, the value of the speech speed magnification N is not changed.

In the loop processing that returns to step 52 via step S52 →... → step S64 as described above, if the input audio signal is a speech signal in which the speech is slowly spoken, the above-described loop processing is repeated.
At 58, the integrated value M exceeds the second threshold value S2 set by the threshold value setting unit 5, and the process proceeds to step S61. In step 61, the speech speed magnification N in Table 1 is shifted right by one from the current value. As a result, the expansion rate of the time axis of the input audio signal becomes slightly smaller than before, and the speech speed becomes slightly faster than before.

Conversely, if the input audio signal is a fast-speech audio signal, the first integrated value M set by the threshold setting unit 5
Since it is smaller than the threshold value S1, the process proceeds from step S58 to step 59. In step 59, the speech speed magnification N in Table 1 is shifted left by one from the current value. As a result, the expansion rate of the time axis of the input audio signal becomes slightly larger than before, and the speech speed becomes slightly lower than before.

As described above, the speech speed magnification N is also changed depending on the storage amount k of the memory.

When a stop instruction is given by the user, the apparatus stops (S64).

In summary, in this embodiment, when the integrated value M is smaller than the first threshold value S1, the expansion rate in the [slow playback mode] is automatically selected so as to be slightly larger, and the integrated value M is set to the second value. Threshold value S2 (however, the first threshold value S
1 <second threshold value S2), the expansion rate in the [slow playback mode] is automatically selected to be slightly smaller, and when the integrated value M is between the first threshold value S1 and the second threshold value, In the slow playback mode].

Next, a fourth embodiment of the present invention will be described with reference to the flowchart of FIG. The basic configuration of the device is the same as that shown in FIGS.
Although the detailed description is omitted, in the present embodiment, the third
Similarly to the embodiment, the threshold setting unit 5 sets the first threshold S
1 and a second threshold value S2 set to a larger value.

In addition, as shown in Table 2 below,
In addition to the [slow playback mode] shown in Table 1, [early playback mode] is added. This [Early Playback Mode] is also the same as [Slow Playback Mode], and is 4-4.
It has two modes.

[0118]

[Table 2]

In FIG. 7, first, the user operates the operation button (not shown) of the mode selection section 8 to select a mode as in the above embodiment (S71). In this embodiment, since there are a [slow playback mode] and a [early playback mode], one of the four modes is selected for each mode. Here, as an example, it is assumed that the [slow playback mode 1] shown in Table 1 and the [early playback mode 1] shown in Table 2 are selected. As a result, the speech speed conversion unit 7 is
Is given to the speech speed control unit 13, and the speech speed control unit 13 is given setting information for the [slow playback mode 1] and the [early speech playback mode 1] based on the given mode selection signal.

Next, the storage amount detector 17 checks the storage amount in the memory 15 to determine the storage amount j, and gives the value of the storage amount j to the speech speed controller 13 (S72). Unlike this, at this point, 1.0 [double speed] (that is, [normal reproduction mode]) is set as the speech speed magnification N as an initial value (S73).

The contents stored in the pitch period storage unit 2 (hereinafter, referred to as a pitch period Tn-1 as a variable name) are initialized (cleared).

Thus, the speech speed magnification N (=
1.0 [double speed]), the pitch period extracting unit 1
Extracts the pitch cycle of the input audio signal (hereinafter, referred to as pitch cycle Tn as a variable name) based on the above equation (S74). As an example, for example, pitch period Tn = 6
It is assumed that 0 [sample] is obtained.

The pitch cycle Tn extracted by the pitch cycle extracting section 1 is compared with the pitch cycle Tn-1 stored in the pitch cycle storing section 2 by the pitch cycle comparing section 3 (S7).
5).

However, since the contents stored in the pitch period storage unit 2 have been cleared as described above, step S8 is executed.
Proceeding to 2, the count value C of the counter 4 is initialized (cleared) to 0, and in step S80 [normal reproduction mode]
Is selected (since [Normal playback mode] has already been set in step S73, there is substantially no mode change).

Further, the value of the pitch period Tn (= 60 [sample]) is stored in the pitch period storage unit 2 and becomes a new value of the pitch period Tn-1 (S83).
Thereafter, the flow returns to step S72.

Then, the storage amount j of the memory is checked again (S72), and the speech speed magnification N based on the storage amount j is set again (S73).

Then, it is assumed that the pitch period Tn is extracted again (S74), and for example, the pitch period Tn = 61 [samples] is obtained. This extracted pitch cycle Tn
(= 61 [sample]) and the pitch period Tn-1 (= 60 [sample]) stored in the pitch period storage unit 2 are compared by the pitch period comparison unit 3 (S75).

Here, the newly extracted pitch period Tn
(= 61 [sample]) and the pitch cycle Tn-1 (= 60 [sample]) stored in the pitch cycle storage unit 2
If the condition of Tn ≒ Tn-1 is satisfied, the count value C of the counter 4 is incremented by one (S56).

Then, the pitch period Tn (= 61 [sample]) extracted by the pitch period extracting unit 1 and the count value C (= 1) of the counter 4 are integrated to obtain an integrated value M (= 61) (S77). Are given to the comparison unit 6.

This embodiment also has a first threshold value S1 and a second threshold value S2 set to a value larger than the first threshold value, as in the third embodiment. When the integrated value M of the pitch period Tn of the input voice signal and the count value C of the counter 4 is between the first threshold value S1 and the second threshold value S2, it is determined that the voice has a standard voice speed. Judge, integrated value M
Is smaller than the first threshold value S1, it is determined that the voice is a fast-talking voice, and if the integrated value M is larger than the second threshold value S1, it is determined that the voice is a voice when the user speaks slowly.

The comparing section 6 includes the integrated value M and the threshold setting section 5
Are compared with the first threshold value S1 (for example, 1000) and the second threshold value S2 (for example, 2000) (S78). As described above, since the integrated value M = 61, it is smaller than the first threshold value S1, and thus the [slow playback mode] is selected (S79). Since the [slow playback mode 1] is set as the initial mode in the [slow playback mode] in step S71, the [slow playback mode 1] is set as the mode to be changed next.

As in the above embodiments, the audio signal whose time axis has been expanded by the time axis compression / expansion unit 11 is encoded by the audio encoding unit 14 and once stored in the memory 15. The audio signal is decoded by the audio decoding unit 16 and becomes an output audio signal.

Then, the value of the pitch cycle Tn is newly stored in the pitch cycle storage unit 2 (S83), and the process returns to step S72 via step S84.

In the loop processing that returns to step 72 after step S72 → step S84 as described above, the above-mentioned loop processing is repeated until the integrated value M falls between the first threshold value S1 and the second threshold value S2. , Step S80, but in this case, the value of the speech speed magnification N is not changed.

Step S72 as described above →
In the loop processing returning to step 72 after step S84, if the input audio signal is a speech signal that is slowly spoken, the above-described loop processing is repeated, and then step S
At 78, the integrated value M exceeds the second threshold value S2 set by the threshold value setting unit 5, and the process proceeds to step S81. In the step 81, a "quick reproduction mode" is selected as a mode to be changed next. In the step S71, since the [early reproduction mode 1] is set as the initial mode in the [early reproduction mode], the [early reproduction mode 1] is set as the mode to be changed next.

After setting the mode, the value of the pitch period Tn (= 61 [sample]) is stored in the pitch period storage unit 2 in the same manner as described above.
And becomes a new value of the pitch period Tn-1 (S8
3) The process returns to step S72.

Then, the storage amount j of the memory is checked again (S72), and the speech speed magnification N based on the storage amount j is set again (S73). In the step S73,
The value of the speech speed magnification N in Table 2 is appropriately changed according to the storage amount j of the memory.

Conversely, when the input audio signal is a fast-speech audio signal, the first integrated value M set by the threshold setting unit 5 is used.
Since it is smaller than the threshold value S1, the process proceeds from step S78 to step 79. In step 79, "slow playback mode" is selected as the mode to be changed next. In step S71, since [slow playback mode 1] is set as the initial mode in [slow playback mode], [slow playback mode 1] is set as the mode to be changed next.

After the mode is set as described above, the value of the pitch period Tn is stored in the pitch period storage unit 2 and becomes the new value of the pitch period Tn-1 in the same manner as described above (S83). An audio signal that has not been expanded or decompressed is output from the time axis compression / expansion unit 11 and the audio encoding unit 1
4 and stored in the memory 15 at one end, and then decoded by the audio decoding unit 16 to become an output audio signal. Then, the process returns to step S72.

Then, the storage amount j of the memory is checked again (S72), and the speech speed magnification N based on the storage amount j is set (S73). In the step S73,
According to the storage amount j of the memory, the value of the speech speed magnification N in Table 1 is appropriately changed.

It is to be noted that the speech speed ratio N is also changed depending on the storage amount k of the memory as described above.

When a stop instruction is given by the user, the apparatus stops (S33).

In summary, in this embodiment, when the integrated value M is smaller than the first threshold value S1, the [slow reproduction mode] is automatically selected, and the integrated value M is set to the second threshold value S2.
(However, when the first threshold value S1 <the second threshold value S2), the [early speed reproduction mode] is automatically selected. When the integrated value M is between the first threshold value S1 and the second threshold value, the [normal reproduction mode] is selected. Mode] is automatically selected.

Further, FIG. 8 is provided with a threshold change operation section 18 for the user to change the threshold set by the threshold setting section 5 in the speech speed conversion apparatus shown in FIG. When listening to voice, there is an individual difference in the feeling that it is difficult to hear at a high speed or that it is too slow. Therefore, the user can use the threshold value changing operation section 18 to automatically change the speech speed.
Alternatively, the first threshold value S1 and the second threshold value S2 can be changed and adjusted so as to automatically set a speech speed suitable for the user. The threshold value changing operation unit 18 is provided with a [+] operation key and a [-] operation key, or an [up] operation key and a [do] operation key.
[wn] Various operation keys such as operation keys, and various forms such as a jog dial and a slide lever can be used. In addition, multiple buttons (such as [fast], [slightly fast], [normal], [slightly slow], [slow], etc.)
Thresholds may be set for these, and the user may select from them.

FIG. 9 is a flow chart showing the operation of the fifth embodiment of the present invention. This is a flow chart showing the operation of the speech speed conversion device shown in FIG. Step S24, which is a setting process, is added. In addition, the same portions as those in the processing shown in FIG. 4 are denoted by the same reference numerals, and detailed description thereof will be omitted.

In step 11, the user sets the mode (any of the slow reproduction modes 1 to 4 shown in Table 1).
Is selected, the threshold value S can be changed by operating the threshold value changing operation unit 18 in the next next step 24.

FIG. 10 is a flowchart showing the operation of the sixth embodiment of the present invention. The user changes and sets the threshold value S in the flowchart showing the operation of the speech speed conversion device shown in FIG. Step S44 is added. In addition, the same parts as those in the processing shown in FIG. 5 are denoted by the same reference numerals, and detailed description thereof will be omitted.

In step 31, the user sets the mode (any of the slow reproduction modes 1 to 4 shown in Table 1).
After selecting, the threshold value S can be changed by operating the threshold value changing operation unit 18 in the next next step 44.

Next, FIG. 11 is a flowchart showing the operation of the seventh embodiment of the present invention. The flowchart showing the operation of the speech speed conversion device shown in FIG.
And step S64, which is a process in which the user changes / sets the second threshold value S2. FIG.
The same reference numerals are given to the same portions as the processes shown in (1), and the detailed description thereof is omitted.

In step 51, the user sets the mode (any of the slow reproduction modes 1 to 4 shown in Table 1).
After the selection of, the threshold change operation unit 18 can be operated in the subsequent next step 64 to change the first threshold S1 and the second threshold S2.

FIG. 12 is a flowchart showing the operation of the eighth embodiment of the present invention. The flowchart showing the operation of the speech speed converter shown in FIG.
Step S85, which is a process in which the user changes and sets the first and second threshold values S2, is added. Further, the same parts as those in the processing shown in FIG. 7 are denoted by the same reference numerals, and detailed description thereof will be omitted.

At step 71, the user sets the mode (one of the slow reproduction modes 1 to 4 shown in Table 1 above,
Alternatively, after selecting any of the fast playback modes 1 to 4 shown in Table 2 above, in the next step 85, the first threshold value S1 and the second threshold value S2 are changed by operating the threshold value changing operation unit 18. Can be.

In each of the above embodiments, the mode selection and threshold setting processes are present only at the beginning of the flowchart for the sake of drawing. However, these processes can be changed as needed during the speech speed conversion process. You may. further,
In each of the above embodiments, both the first threshold value and the second threshold value are configured to be changed, but one of them may be configured to be changed.

[0154]

As described in detail above, according to the present invention, it is determined whether or not an input audio signal is an audio signal when speaking in a utterance, and only for an audio signal when speaking in an utterance. Since it can be converted into a slow voice signal by voice speed conversion processing, if the input voice signal is a voice signal that is slowly spoken, the voice speed conversion processing is performed and the voice speed is converted to a voice signal that is further slowly spoken Nothing.

Further, according to the present invention, the magnification of the speech speed conversion processing is appropriately changed according to the speech speed.

Further, according to the present invention, not only the scaling factor of the speech speed conversion process is appropriately changed according to the speech speed, but also the speech speed conversion process is performed for a speech signal spoken at a standard speed. Not done.

[0157] According to the present invention, the speech rate conversion process is performed on the fast-speech audio signal so as to convert the speech signal into a slowly spoken speech signal. The speech speed conversion process is performed in order to achieve the above, and the speech speed conversion process is not performed on the voice signal spoken at a standard speed.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram showing a configuration of a speech speed conversion device of the present invention.

FIG. 2 is a schematic block diagram showing a configuration of a speech speed conversion unit in the speech speed conversion device of the present invention.

FIG. 3 is a diagram for explaining the operation of the speech speed conversion device of the present invention.

FIG. 4 is a flowchart for explaining the operation of the speech speed conversion device of the present invention.

FIG. 5 is a flowchart for explaining the operation of the second embodiment of the present invention.

FIG. 6 is a flowchart for explaining the operation of the third embodiment of the present invention.

FIG. 7 is a flowchart for explaining the operation of the fourth embodiment of the present invention.

FIG. 8 is a schematic block diagram showing a configuration of a speech speed conversion device according to another embodiment of the present invention.

FIG. 9 is a flowchart for explaining the operation of the fifth embodiment of the present invention.

FIG. 10 is a flowchart for explaining the operation of the sixth embodiment of the present invention.

FIG. 11 is a flowchart for explaining the operation of the seventh embodiment of the present invention.

FIG. 12 is a flowchart for explaining the operation of the eighth embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Pitch period extraction part 2 Pitch period storage part 3 Pitch period comparison part 4 Counter 5 Threshold setting part 6 Comparison part 7 Speech rate conversion part 8 Mode selection part 11 Voice time axis compression / decompression part 12 Silence section detection part 13 Speech rate control part 14 audio encoding unit 15 memory 16 audio decoding unit 17 accumulated amount detecting unit 18 threshold value changing operation unit

Claims (16)

[Claims] 1. A pitch cycle detecting means for detecting a pitch cycle from an audio signal; a counting means for counting the number of repetitions of a predetermined pitch cycle based on the pitch cycle extracted by the pitch cycle detecting means; Comparing and judging means for comparing a product of the pitch period extracted by the means and the number of repetitions counted by the counting means with a predetermined threshold; and speech speed converting means for performing speech speed conversion based on the judgment result of the comparing and judging means And a speech speed conversion device. 2. A pitch cycle detecting means for detecting a pitch cycle from an audio signal; a counting means for counting the number of repetitions of a predetermined pitch cycle based on the pitch cycle extracted by the pitch cycle detecting means; Comparing and judging means for comparing a product of the pitch period extracted by the means and the number of repetitions counted by the counting means with a predetermined threshold; threshold changing means for changing the predetermined threshold; judgment results of the comparing and judging means And a speech speed conversion means for performing speech speed conversion based on the speech speed. 3. The speech speed converting means, wherein the product of the pitch cycle extracted by the pitch cycle detecting means and the number of repetitions counted by the counting means does not exceed a predetermined threshold value in the comparing and judging means. 3. The speech speed conversion device according to claim 1, wherein the speech speed is converted into a slow voice signal. 4. The speech speed conversion means, when the comparison determination means determines that the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means does not exceed a predetermined threshold. The voice signal is converted into a slow voice signal and the speech speed is converted, and the product of the pitch cycle extracted by the pitch cycle detection means in the comparison determination means and the number of repetitions counted by the counting means exceeds a predetermined threshold value. 4. The speech speed conversion device according to claim 1, wherein the speech speed conversion is not performed when the judgment is made. 5. The speech speed conversion means, when the comparison / determination means determines that the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means is within a predetermined threshold value. If the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means is determined to be greater than a predetermined threshold, the speech rate magnification is decreased. 4. The speech speed conversion device according to claim 1, wherein the speech speed conversion is performed. (However, speech speed ratio = time length of input voice signal / time length of output voice signal) 6. A pitch cycle detecting means for detecting a pitch cycle from an audio signal; a counting means for counting the number of repetitions of a predetermined pitch cycle based on the pitch cycle extracted by the pitch cycle detecting means; Means for comparing a product of the pitch period extracted by the means and the number of repetitions counted by the counting means with a predetermined first threshold value and a predetermined second threshold value; And a speech speed conversion means for performing speech speed conversion. (However, the first threshold value <the second threshold value) 7. A pitch cycle detecting means for detecting a pitch cycle from an audio signal; a counting means for counting the number of repetitions of a predetermined pitch cycle based on the pitch cycle extracted by the pitch cycle detecting means; Means for comparing a product of the pitch period extracted by the means and the number of repetitions counted by the counting means with a predetermined first threshold value and a predetermined second threshold value; and the predetermined first threshold value or the predetermined value. A speech rate conversion device comprising: a threshold value changing unit that changes the second threshold value; and a speech speed conversion unit that performs speech speed conversion based on the determination result of the comparison determination unit. (However, the first threshold value <the second threshold value) 8. The speech speed conversion means, wherein the product of the pitch cycle extracted by the pitch cycle detection means in the comparison determination means and the number of repetitions counted by the counting means is a predetermined first threshold value and a predetermined first threshold value. 8. The speech speed conversion device according to claim 6, wherein the speech speed conversion is not performed when it is determined that the speech speed is between two threshold values. 9. The speech speed conversion means, wherein the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means is smaller than a predetermined first threshold value in the comparison determination means. 9. The speech speed conversion device according to claim 6, wherein when the judgment is made, the speech speed conversion is performed with the speech speed magnification reduced. (However, speech speed ratio = time length of input voice signal / time length of output voice signal) 10. The speech speed conversion means, wherein the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means in the comparison determination means is greater than a predetermined second threshold value. 10. The speech speed conversion device according to claim 6, wherein when the judgment is made, the speech speed conversion is performed by increasing the speech speed magnification. (However, speech speed ratio = time length of input voice signal / time length of output voice signal) 11. The speech speed conversion means, wherein the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means in the comparison determination means is smaller than a predetermined first threshold value. 9. The speech speed conversion device according to claim 6, wherein when the judgment is made, the speech speed conversion is performed with the speech speed magnification smaller than one. (However, speech speed ratio = time length of input voice signal / time length of output voice signal) 12. The speech speed conversion means, wherein the product of the pitch cycle extracted by the pitch cycle detection means and the number of repetitions counted by the counting means in the comparison determination means is greater than a predetermined second threshold value. 9. The speech speed conversion device according to claim 6, wherein when the judgment is made, the speech speed conversion is performed with the speech speed magnification greater than 1. (However, speech speed ratio = time length of input voice signal / time length of output voice signal) 13. The speech speed conversion unit according to claim 1, wherein the speech speed conversion unit changes the speech speed magnification in accordance with an empty capacity of the storage unit that stores the speech signal whose speech speed has been converted. Speech speed converter. (However, speech speed ratio = time length of input voice signal / time length of output voice signal) 14. The speech speed conversion means changes the speech speed magnification to be closer to 1 as the empty capacity of the storage means for storing speech speed converted speech signals decreases. The speech speed conversion device according to claim 1. (However, speech speed ratio = time length of input voice signal / time length of output voice signal) 15. The speech speed conversion means changes the speech speed magnification to be closer to a predetermined magnification in accordance with an increase in the empty capacity of the storage means for storing the speech speed converted speech signal. 14. The speech speed conversion device according to claim 1, wherein: (However, speech speed ratio = time length of input voice signal / time length of output voice signal) 16. The pitch cycle according to claim 1, wherein the predetermined pitch cycle is the same pitch cycle, a double pitch cycle, a half pitch cycle, and a pitch cycle that is close to them. Item 16. The talking-side conversion device according to Item 15.