JP2001154684A - Speech speed converter - Google Patents

Abstract

PROBLEM TO BE SOLVED: To gradually change the speech speed of the beginning of the speech to a target speech speed in a speech speed converter. SOLUTION: The converter analyzes inputted digital voice signals, synthesizes the analyzed digital voice signals, generates new voice signals having a specified speech speed, converts the new voice signals into analog voice signals and outputs the signals to the external. The converter is provided with a sound detecting means 13 which detects a first sound in the inputted voice signals and a speech speed computing section 14 which changes a normal speech speed to a faster target speech speed within a prescribed time interval starting from the sound detecting time of the means 13, successively computes a speech speed to maintain the target speech speed and transmits the speech to a signal synthesis section.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech speed conversion device for changing the speech speed of voice.

[0002]

2. Description of the Related Art For example, as a method of learning a foreign language, a learning method of actually learning by listening to the foreign language by ear is effective. In this case, it is important to listen to the same conversation repeatedly. Therefore, in order to improve the ability to listen to a foreign language, as a simple method, for example, learning is performed by playing a tape on which a conversation or reading for several minutes to several tens of minutes is recorded.

[0003] In this case, a tape recorder dedicated to language learning is configured so that the sound reproduction speed can be varied within a certain range. Therefore, the beginner learner can listen to the recorded conversation at a slower playback speed, and the advanced learner can improve the learning efficiency by listening to the recorded conversation at a faster playback speed.

However, simply changing the reproduction speed of the sound changes the frequency of the reproduced sound, and the sound sounds unnatural.

[0005] Even if the speaking speed, that is, the speaking speed is changed in order to solve such inconvenience, the frequency of the voice does not change, and the sound is heard naturally only by slowing down the speech or making the speaker speak faster. A speech speed conversion method has been proposed. Next, this speech speed conversion method will be described with reference to FIGS.

FIG. 13 shows, for example, "It's difficult for
It is a waveform diagram of the audio signal 1 when saying "me to finish ...". FIG. 14 is an enlarged view of the audio signal 1. As is well known, speech includes consonants and vowels, and audio signal 1
Also have corresponding consonants and vowels. As shown, the consonant is composed of one unvoiced sound 2 and the vowel is composed of a plurality of voiced sounds 3. The audio signal 1 includes silence 4 when the audio is interrupted.

[0007] The unvoiced sound 2 constituting a consonant has a relatively high frequency component, and the plurality of voiced sounds 3 constituting a vowel have substantially the same waveform. Therefore, in order to increase the speech speed, one or a plurality of voiced sounds 3 among the plurality of voiced sounds 3 constituting the vowel are thinned out, and the voiced sounds 3 before and after the thinned voiced sound 3 or unvoiced sounds 3 are mixed. 2 and voiced sound 3 or voiced sound 3
And silence 4 are connected. Therefore, the duration of the vowel can be shortened, and as a result, the entire time of the audio signal 1 can be shortened,
Speaking speed can be increased without changing the frequency or sound quality of the voice.

Conversely, when the speech speed is to be reduced, the same voiced sound 3 may be inserted into a plurality of voiced sounds 3 constituting a vowel to extend the duration of the vowel.

In order to automatically perform the speech speed conversion, it is necessary to classify the unvoiced sound 2, the voiced sound 3 and the silent sound 4 included in the voice signal 1. As a classification method, a vowel uses the property that a plurality of voiced sounds 3 continue, and calculates an autocorrelation function for the audio signal 1, thereby classifying the unvoiced sound 2 and the voiced sound 3 and each voiced sound. The duration (pitch) of 3 is detected.

[0010] The speech speed of the voice signal 1 is determined by how many voiced sounds 3 among a plurality of voiced sounds 3 constituting a vowel are thinned out or how many voiced sounds 3 are inserted.

[0011] Therefore, by using a voice reproducing apparatus incorporating such a speech speed conversion function, a language learner can listen to conversation or narration at a high speed or a low speed without a sense of incongruity.

[0012]

However, even in a sound reproducing apparatus incorporating the above-mentioned speech speed conversion function,
There were the following issues that still need to be improved.

That is, in the audio reproducing apparatus in which the above-mentioned speech speed conversion function is incorporated, the speed is the same from the start of the reproduction of the sound to the end of the reproduction. Therefore, even for advanced learners, even if the content is played back immediately after the start of playback, it takes time to get used to the playback, and there is a problem that the beginning of a conversation or narration cannot be heard accurately.

[0014] The present invention has been made in view of such circumstances, and the speech speed is normal at the beginning, and then gradually increased, so that even if the speech speed is set to be high, the conversation or An object of the present invention is to provide a speech speed conversion device that can accurately recognize the beginning of a narration and that is easy for a language learner to use.

[0015]

In order to solve the above-mentioned problems, the present invention provides an audio signal memory for storing an input digital audio signal, and a digital audio signal output from the audio signal memory. The present invention is applied to a speech speed conversion device comprising a voice analysis unit for performing the above-described processing, and a signal synthesis unit for synthesizing the digital voice signal analyzed by the voice analysis unit and outputting as a new voice signal having a specified voice speed. You.

In order to solve the above-mentioned problem, the present invention provides a sound detecting means for detecting the first sound in an input audio signal, and a specified time from a voice sound detecting time of the sound detecting means. Within the period until the elapsed time, the speech speed is changed to the target speech speed higher than the normal speech speed with the passage of time, and after the lapse of the specified time, the speech speed for maintaining the target speech speed is sequentially calculated and transmitted to the signal synthesis unit. And a speech speed calculation unit.

In the thus constructed speech speed conversion device, the input speech signal is temporarily stored in the speech signal memory, read out, and then read into a new speech signal having the speech speed designated by the signal synthesis unit. Is synthesized. Then, the speech speed applied to the signal synthesizing unit is increased from the detection time of the first sound in the audio signal to the target speech speed higher than the normal speech speed with the lapse of time from the detection time of the first sound to the specified time, Thereafter, control is performed so as to maintain the target speech speed.

Therefore, in the reproducing apparatus in which the speech speed conversion device is incorporated, only the beginning portion of the conversation or the narration becomes the normal speech speed, and thereafter the target speech speed becomes faster than the normal speech speed. Thus, language learners do not miss the beginning of a conversation or narration.

Further, another invention relates to an audio signal memory for storing an input digital audio signal, and an audio analysis for dividing the digital audio signal output from the audio signal memory into unvoiced sound, voiced sound, and silent sound. Unit and only the voiced sound segmented by the speech analysis unit are decimated in accordance with the specified speech speed, and the unvoiced sound and the decimated voiced sound and silence are connected to synthesize a new speech signal. The present invention is applied to a speech speed conversion device including a signal synthesis unit.

In order to solve the above-mentioned problem, according to the present invention, a target speech speed higher than a normal speech speed with a lapse of time within a period from a start time of an input audio signal to a lapse of a specified time. And a speech speed calculation unit for sequentially calculating a speech speed for maintaining a target speech speed after a lapse of a specified time and sending the calculated speech speed to a signal synthesis unit.

In the thus constructed speech speed converter, the input speech signal is temporarily stored in the speech signal memory and then divided by the speech analysis unit into the above-mentioned unvoiced sound, voiced sound and silence. . Then, in the next signal synthesizing unit, the number of voiced sounds corresponding to the speech speed higher than the designated normal speech speed is thinned out to become a new voice signal of the designated speech speed.

The speech speed to be supplied to the signal synthesizing unit is, in the speech speed calculating unit, a period from the start time of the input voice signal to a lapse of a predetermined time, and a target speed higher than the normal speech speed with time. It is controlled to increase to the speech speed and then maintain the target speech speed.

Therefore, it is possible to obtain substantially the same effect as the above-described invention.

According to another aspect of the present invention, in the above-described speech speed conversion device, the start time of the input voice signal is the start time of the first voiced sound of the voice signal which is divided by the voice analysis unit.

As described above, voice signals include unvoiced sounds corresponding to consonants and voiced sounds corresponding to vowels. "Shhh"
Unvoiced sounds such as "chic" and "chee" do not make much sense, and voiced sounds often make sense. Therefore, it is desirable to start changing the speech speed from the start time of the voiced sound.

In another aspect of the present invention, the speech rate multiplying unit in the speech rate converting apparatus described above is characterized in that the speech analyzing unit measures a duration of the silence from a silence start time for each silence in the speech signal. A timer is provided, and a speech speed discontinuity calculator is configured to calculate a speech speed that increases in accordance with the number of times each time the silent duration measuring unit measures the threshold time.

In the thus constructed speech speed conversion device, the speech speed increases each time a silent period equal to or greater than a threshold value occurs during conversation or narration. Therefore, it is possible to prevent the speech speed from being changed in the middle of one word or one sentence, and to listen as a more natural word.

According to still another aspect of the present invention, there is provided the above-mentioned speech speed conversion apparatus, wherein the voice analysis unit measures the duration of the corresponding silence from the silence start time at each silence in the voice signal; Every time the period timer counts the threshold time, it changes from the start time of the unvoiced sound or voiced sound that follows the corresponding silence to a target voice speed higher than the normal voice speed with the passage of time, and thereafter the target voice speed is changed. A speech speed calculating unit for sequentially calculating the maintained speech speed and sending it to the signal combining unit.

In the thus constructed speech speed converter, the speech speed is changed from the normal speech speed to the normal speech speed every time a silent period equal to or greater than a threshold value occurs during a conversation or narration. It changes to a fast target speech speed.

[0030]

Embodiments of the present invention will be described below with reference to the drawings. (First Embodiment) FIG. 1 is a block diagram showing a schematic configuration of a speech speed conversion device according to a first embodiment of the present invention. A series of audio signals a having the same configuration as the audio signal 1 shown in FIG. Therefore, as shown in FIG. 14, this audio signal a is a voiceless sound 2 corresponding to a consonant.
And a voiced sound 3 corresponding to a vowel and a silence 4. Then, in the speech speed conversion device of this embodiment, the duration of the unvoiced sound 2 or the voiced sound 3 is defined as a voiced period 16, and the duration of the silence 4 is defined as a silent period 17.

The analog audio signal a input from the input terminal 5 is converted into a digital audio signal by the A / D converter 5 and then stored in the audio signal memory 7. The audio analysis unit 8 divides the series of digital audio signals a ₁ written in the audio signal memory 8 into unvoiced sounds 2, voiced sounds 3, and silent sounds 4. Specifically, the speech level of the audio signal a ₁ is checked, and the sound period 16 and the silence period 17 are classified. After that, the autocorrelation analysis is performed on the signal of each voiced period 16 to determine the voiced period 16 as unvoiced sound 2 and voiced sound 3.
And is divided into

In the voice analysis unit 8, the unvoiced sound 2, the voiced sound 3 and the
Sound signal a divided into sound 4_TwoEnters the signal synthesizer 9
Is forced. The signal synthesizing unit 9 receives the input audio signal a_TwoTo
Speed calculation of voiced sounds 3 that make up each vowel
Corresponds to the speech speed magnification Y as the speech speed specified in the output unit 14
Decimate by the number you chose. Then, the input audio signal a_TwoTo
The unvoiced sound 2 and the decimated voiced sound 3 and silence 4
And a new audio signal a _ThreeAnd output. here,
The speech speed magnification Y is a normal speech speed without changing the speed.
Is 1 (reference).

The new audio signal a ₃ output from the signal synthesizing unit 9 is temporarily stored in an output buffer 10, converted into an analog audio signal a ₄ by a D / A converter 11, and output from an output terminal 13. You. Therefore, the new analog audio signal a ₄ output from the output terminal 12 is shorter than the analog audio signal a input to the input terminal 5 by the specified speech speed magnification Y, and Minutes, the speed of the replayed conversation or narration is increased.

The digital audio signal a ₁ output from the audio signal memory 7 is input to the audio analyzer 8 and
It is input to the sound detection unit 13. The sound detection unit 13 checks the signal level of the first audio signal a ₁ in the input audio signal a ₁ , and notifies the speech speed calculation unit 14 of the beginning of the sound period 16 as sound detection.

[0035] The speech speed calculation part 14, with the specified time T _B from the time setting unit 18 is set, the target speech speed setting unit 19
, The target speech speed magnification A as the target speech speed is set.

Next, the speech speed magnification Y in the speech speed calculation unit 14
Will be described. The speech rate calculation unit 14, as shown in FIG. 3, within the period from sound detection time t _S of the digital audio signal a ₂ to the specified time T _B has elapsed time t _E, usually talking speech speed ratio Y Normal speech speed magnification corresponding to speed (Y = 1)
From the target voice speed magnification A higher than the normal voice speed.

Specifically, the processing shown in FIG. 2 is performed.
In FIG. 2, when the sound detection unit 13 detects a sound (R
1) The elapsed time T is initialized (T = 0) (R2). When the short time Δt has elapsed (R3), the elapsed time T is updated at R4 (T = T + Δt). When the elapsed time T after the update is less than the predetermined time T _B (R5), and calculates the speech speed ratio Y shown in the following formula (R6). Y = [(A-1) / T _B ] T + 1 The calculated speech speed magnification Y is transmitted to the signal synthesizing unit 9 (R
7). Then, the process returns to R3 and waits for the elapse of the next minute time Δt.

[0038] At R5, the elapsed time T after the update reaches the predetermined time T _B, and sends the target speech speed ratio Y = A to the signal synthesizing unit 9.

In the speech speed conversion device thus configured, as shown in FIG. 3, when the input of the audio signal a is started at time t ₀ , the audio signal a output from the output terminal 12 is output.
The voice speed of ₄ is the normal voice speed (Y = 1). Then, the time t _S
At, the sound period 16 starts, and when sound is detected,
Timing of the elapsed time T is started, and the speech speed magnification Y starts increasing.

[0040] Then, at time t _E the elapsed time T has exceeded the specified time T _B from the time t _S, speech speed ratio Y reaches the target speech speed ratio A (Y = A). Time t _E after a lapse of specified time T _B is speaking rate ratio Y maintains the target speech speed ratio A.

Therefore, by using this speech speed conversion device, only the beginning portion of the conversation or narration becomes the normal speech speed, and thereafter the target speech speed becomes faster. Thus, language learners do not miss the beginning of a conversation or narration.

(Second Embodiment) FIG. 4 is a block diagram showing a schematic configuration of a speech speed conversion device according to a second embodiment of the present invention. The same portions as those of the speech speed conversion device of the first embodiment shown in FIG. 1 are denoted by the same reference numerals, and detailed description of the overlapping portions will be omitted.

In this speech speed conversion device, the speech signal a ₂ divided into the unvoiced sound 2, the voiced sound 3 and the silence 4 by the speech analysis unit 8 is input to the signal synthesis unit 9 and the speech speed calculation unit 14a. . Signal combining unit 9, among the plurality of voiced 3 constituting each vowel in the speech signal a ₂ input, thins out the number corresponding to the designated speech rate ratio Y in the speech speed setting unit 14a. Then, the unvoiced sound ₂ in the input audio signal a ₂ and the voiced sound 3 and the silence 4 after the thinning are connected to synthesize a new audio signal a ₃ and output it.

Therefore, the speech speed of the new analog voice signal a ₄ output from the output terminal 12 is specified by the voice speed setting unit 14 a with respect to the analog voice signal a input to the input terminal 1. It becomes the speaking speed of the speaking speed magnification Y.

The speech speed calculation unit 14a receives the number of steps M from the step number setting unit 20 and the target speech speed magnification A from the target speech speed setting unit 19. The number of steps M is the number of steps at which the voice speed Y is shifted from the normal voice speed (Y = 1) to the target voice speed A.

Next, the operation of calculating the speech speed magnification Y in the speech speed calculator 14a will be described. This speech speed calculator 14a
As shown in FIG. 6, in the period until the specified time T _B has elapsed from the input time t _S of the audio signal a _2, faster target than normal speaking speed the speech speed ratio Y from the normal speech speed (Y = 1) M to speech speed magnification A
Perform the process of changing in steps.

Specifically, the processing shown in FIG. 5 is performed.
In FIG. 5, the number of steps I of the speech speed magnification Y in FIG.
Is set to an initial value 0 (I = 0), and a time lapse flag F indicating that the duration T of the silent period 17 of the audio signal a ₂ in FIG. 7 exceeds the threshold time T _S is set to the initial value 0. Setting (F
= 0) (S1).

When the minute time Δt has elapsed (S2), the analysis result of the audio signal a ₂ at that time is read (S2).
3). When the analysis result is the silent period 17 (S
4) In S5, the duration T ₁ of the silent period 17 is updated (T ₁ = T ₁ + Δt).

The updated duration T _! Is equal to the threshold time T _S
Does not elapse (S6), the process returns to S2, and waits for the elapse of the minute time Δt. If the updated duration T ₁ has exceeded the threshold time T _S (S6), the duration T ₁ of the silence period 17 is cleared to 0 (S7), and the time elapsed flag F
Remains at 0 (S8), the duration T ₁ has exceeded the threshold time T _S for the first time in the silence period 17, so the time lapse flag F is set to 1 (S9). And
In S10, the number of steps I is updated (I = I + 1).

Then, when it is confirmed that the number of steps I does not exceed the final number of steps M corresponding to the target speech speed magnification A corresponding to the speech speed higher than the normal speech speed (S11), the speech speed magnification is calculated by the following equation. Y is calculated (S12). Y = [(A-1) / M] I + 1 The calculated speech speed magnification Y is sent to the signal synthesizing unit 9 (S1).
3).

At S8, the time lapse flag F has already been set.
Is set to 1, since the calculation (updating) of the speech speed magnification Y has already been completed within the silent period 17, the process returns to S2 without doing anything.

Further, in S4, instead of the silent period 17,
In the case of the sound period 16, the process proceeds to S 14, where the duration T ₁ of the silence period 17 is cleared to 0, and the time excess flag F is cleared.

Further, if the number of steps I exceeds the final number of steps M corresponding to the target speech speed magnification A in S11, the speech speed has already reached the target speech speed magnification A higher than the normal speech speed. Return to S2 without any action.

In the speech speed conversion device of the second embodiment thus configured, the speech signal a ₂ divided into the unvoiced sound 2, the voiced sound 3 and the silence 4 having the signal waveform shown in FIG. If input to the calculation unit 14a, the time t ₂ ~t ₃ sections, the time t ₄ ~t ₆ intervals, and in the period from time t ₆ ~t _8, although silent period 17 exists, the second time t ₂ ~
only the duration T of t ₃ of the section is, at time t _5, since passed the threshold time T _S, at this time t _5, the speech speed magnification Y
Is updated by one step in the direction in which the speech speed increases.

Therefore, as shown in FIG. 6, in the input voice signal “a”, from the input start time of the voice signal,
Each time a silent period 17 longer than the threshold time T _S occurs,
The speech speed is changed until the speech speed reaches the target speech speed magnification A. When the speech speed reaches the target speech speed magnification A, the target speech speed magnification A is maintained thereafter.

Therefore, similarly to the speech speed conversion device of the first embodiment shown in FIG. 1, by using the speech speed conversion device of the second embodiment, only the beginning portion of conversation or narration is changed to the normal speech speed and thereafter Becomes the target speech speed. Thus, language learners do not miss the beginning of a conversation or narration.

Further, in the speech speed conversion device of the second embodiment, each time a silent period 17 that is present in the middle of a conversation or a narration and has a threshold time T _S or more occurs, the speech speed magnification Y is increased. The speed is changed in the increasing direction. Therefore, it is possible to prevent the speech speed from being changed in the middle of one word or one sentence, and to listen as a more natural word.

FIG. 8 shows the actual conversation speed (narration) from the normal speech speed (Y = 1) to the intermediate speech speed (1 <Y <A).
From a diagram showing the waveform of the audio signal a ₄ output from the output terminal 12 in the case of changing to the target speech speed (Y = A). As shown in the figure, it can be understood that when the voice speed magnification Y increases, a plurality of voiced sounds 3 constituting each vowel are thinned out.

(Third Embodiment) Next, a speech speed converter according to a third embodiment of the present invention will be described. The hardware configuration of the speech speed conversion device of the third embodiment is the same as that of the speech speed conversion device of the second embodiment shown in FIG. The only difference is the software configuration of the speech speed calculation unit 14a.

That is, the speech speed calculator 14a in the speech speed converter of the third embodiment, as shown in FIG.
Every time the duration T ₁ of the silent period 17 in the audio signal a ₂ exceeds the threshold time T _S , from the start time of the voiced period 16 consisting of the unvoiced sound 2 or voiced sound 3 following the relevant silent period 17. Normal speech speed (Y = 1) as the elapsed time T increases
To the target speech speed M which is higher than the normal speech speed, and thereafter the speech speed Y which keeps the target speech speed M is sequentially calculated and sent to the signal synthesizing unit 9.

Specifically, the processing shown in FIG. 9 is performed.
In FIG. 9, a time lapse flag F indicating that the duration T ₁ of the silence period 17 of the audio signal a ₂ in FIG. 10 exceeds the threshold time T _S is set to an initial value 0 (F = 0) (FIG. 10). Q1).

When the short time Δt has elapsed (Q2), the analysis result of the audio signal a ₂ at that time is read (Q2).
3). When the analysis result is the silent period 17 (Q
4), in the Q5, to update the duration T ₁ of the silent period _{17 (T 1 = T 1 +} Δt).

If the updated continuation time T ₁ has not exceeded the threshold time T _S (Q 6), the flow returns to Q 2 and waits for the lapse of the minute time Δt. If the updated duration T ₁ has exceeded the threshold time T _S (Q6), the duration T _{1 is set} to 0.
(Q7), and if the time lapse flag F remains 0 (Q8), the duration T ₁ has exceeded the threshold time T _S for the first time in the silence period 17, so that the time lapse flag F Is set to 1 (Q9). And in Q10,
The elapsed time T for changing the voice speed magnification Y is initialized to 0 (T = 0).

If the time lapse flag F has already been set to 1 at Q8, the process returns to Q2 without doing anything.
In Q4, in the case of the silent period 17, the elapsed time T is updated in Q11 (T = T + Δt), and the time elapsed flag F is set to 1
Is set to (Q12), the elapsed time flag F is cleared to 0, and the elapsed time T is cleared to 0 (Q13). Then, the process proceeds to Q14. If the time lapse flag F has already been cleared to 0 in Q12, the process directly proceeds to Q14.

In Q14, the speech speed magnification Y is calculated using the following equation.

Y = A− (A−1) exp [−T / (T _B / 5)] The calculated speech speed magnification Y is set in the signal synthesizing unit 9 (Q 1
5).

In the speech speed converter according to the third embodiment having the above-described structure, the voiced time 16 composed of the unvoiced sound 2 or the voiced sound 3 having the signal waveform shown in FIG. audio signal a ₂ which is divided into consider the case where input to the speech rate calculation unit 14a.

In this case, during the first sound period 17, the speech speed changes so as to increase to the target speech speed (Y = A). Even if the first silent period 17 arrives, the duration T ₁ of the silent period 17 is shorter than the threshold time T _S ,
The speech speed does not return. Then, the speech speed continues to change in the first sound period 16 and when the speech speed reaches the target speech speed (Y = A), the target speech speed (Y = A) is maintained.

When the second silence period 17 arrives, the duration T ₁ of the silence period 17 is longer than the threshold time T _S , so that the speech speed returns to the original speed at the start of the next third speech period 16. . Then, the speech speed starts changing again so as to increase to the target speech speed (Y = A).

Therefore, as shown in FIG. 11, in the entire section from the start time to the end time of the series of voice signals a, the speech speed changes from the normal speech speed (Y = 1) to the target speech speed (Y = A).
Repeat the change to.

Therefore, every time a silent period longer than the threshold time that occurs during conversation or narration occurs,
Since the speech speed changes from the normal speech speed to the target speech speed which is higher than the normal speech speed, it is possible to achieve substantially the same operation and effect as the speech speed conversion device of the first embodiment described above.

The present invention is not limited to the above embodiment. In the speech speed conversion devices of the first, second, and third embodiments, as shown by the D characteristic in FIG. 12, the speech speed changes from the normal speech speed (Y = 1) to the target speech speed (Y =
The timing for starting the change to A) is determined by the audio signal a.
The start position of the first unvoiced sound 2 in the sound period 16 of (a ₂ ) was set.

However, as shown by the E characteristic in FIG. 12, the speech speed changes from the normal speech speed (Y = 1) to the target speech speed (Y = A).
It is also possible to set the timing to start changing to the start position of the voiced sound 3 at the beginning of the voiced period 16 of the audio signal a (a ₂ ).

The unvoiced sound 2 such as "sheep" or "chee" at the beginning of the voice does not make much sense, and the voiced sound 3 often makes sense. Therefore, E in FIG.
As shown in the characteristics, it is desirable to start changing the speech speed from the start time of the voiced sound 3.

The present invention is not limited to the above embodiments. In the speech speed converter of each embodiment, as shown in FIGS. 1 and 4, an A / D converter 6 is provided to convert an analog audio signal input from the input terminal 5 into a digital audio signal. To be written into the audio signal memory 7. Further, the digital audio signal output from the output buffer 10 and subjected to the speech speed change processing is converted into an analog audio signal using the D / A converter 11 and output from the output terminal 12.

However, when a speech speed change process is performed on a digital voice signal that has already been digitally processed by a computer or the like, or when the digital voice signal after the speech speed change process is digitally processed again by a computer or the like, Inputs a digital audio signal directly from the input terminal 5a to the audio signal memory 7 as shown by a chain line in FIGS. Also, the digital audio signal output from the output buffer 10 and subjected to the speech speed change processing is directly output to the output terminal 12a.

[0077]

As described above, in the speech speed conversion apparatus according to the present invention, the speech starts at the normal speech speed, and the speech speed is gradually increased to the target speech speed by the lapse of the specified time. I have to.

Therefore, even if the speaking speed is set to be high, the beginning of the conversation or the narration can be accurately distinguished, and the language learner can easily use the language learning device incorporating the speech speed conversion device. Can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of a speech speed conversion device according to a first embodiment of the present invention;

FIG. 2 is a flowchart showing a calculation process of a speech speed magnification of a speech speed calculation unit incorporated in the same speech speed conversion device.

FIG. 3 is a diagram showing a change in speech speed in the same speech speed conversion device.

FIG. 4 is a block diagram showing a schematic configuration of a speech speed conversion device according to a second embodiment of the present invention;

FIG. 5 is a flowchart showing a calculation process of a speech speed magnification of a speech speed calculation unit incorporated in the same speech speed conversion device.

FIG. 6 is a diagram showing a change in speech speed in the same speech speed conversion device.

FIG. 7 is a diagram showing an audio signal waveform for explaining the operation of the same speech speed conversion device.

FIG. 8 is a diagram showing the transmission of audio signals having different voice speeds output from the output terminal of the voice speed converter.

FIG. 9 is a flowchart showing a process of calculating a speech speed magnification of a speech speed calculation unit incorporated in a speech speed conversion device according to a third embodiment of the present invention.

FIG. 10 is a diagram showing an audio signal waveform for explaining the operation of the same speech speed conversion device.

FIG. 11 is a diagram showing a change characteristic of a voice speed in the voice speed converter.

FIG. 12 is a diagram showing a change characteristic of a speech speed in a modification of the speech speed conversion device.

FIG. 13 is a diagram showing a general audio signal waveform.

FIG. 14 is a diagram showing details of a general audio signal.

[Explanation of symbols]

 2 unvoiced sound 3 voiced sound 4 silent 5 input terminal 6 A / D converter 7 audio signal memory 8 audio analyzer 9 signal synthesizer 10 output buffer 11 D / A converter 12 output Terminal 13: Voice detection unit 14, 14a: Voice speed calculation unit 16: Voice period 17: Silence period 18: Time setting unit 19: Target voice speed setting unit 20: Step number setting unit

Claims (5)

[Claims] 1. An audio signal memory (7) for storing an input digital audio signal, an audio analysis unit (8) for analyzing a digital audio signal output from the audio signal memory, and an audio analysis unit A signal synthesizing unit (9) for synthesizing the digital audio signal analyzed in (1) and outputting it as a new audio signal having a specified speech speed. Voice detection means (13) for detecting the presence of voice, and within a period from the voice detection time of the voice detection means until a lapse of a specified time (T _B ), the voice speed is higher than the normal voice speed with the lapse of time. A speech speed calculation unit (14) for changing the speech speed to a target speech speed and sequentially calculating the speech speed for maintaining the target speech speed after the lapse of the specified time, and sending the speech speed to the signal synthesizing unit. Speaking speed converter. 2. An audio signal memory (7) for storing an input digital audio signal, and an audio analysis unit (11) for dividing the digital audio signal output from the audio signal memory into unvoiced sound, voiced sound, and silent sound. 8) and thinning out only the voiced sounds classified by the voice analysis unit in accordance with the specified speech speed, and connecting the unvoiced sound, the decimated voiced sound and the unvoiced sound to form a new voice signal. And a signal synthesizing unit (9) for synthesizing the normal speech with a lapse of time within a period from a start time of the input audio signal until a lapse of a specified time (T _B ). A speech speed calculation unit (14a) that changes the speech speed to a target speech speed higher than the speed and after the lapse of the prescribed time, sequentially calculates a speech speed that maintains the target speech speed and sends the speech speed to the signal synthesis unit. Characteristic speech speed converter. 3. The start time of the input audio signal is:
3. The speech speed conversion device according to claim 2, wherein the speech signal is a start time of a first voiced sound segmented by the speech analysis unit in the speech signal. 4. The speech speed calculation section, wherein the speech analysis section counts a duration of the silence from a silence start time in each silence in the speech signal, and a silence duration timer; 4. A speech speed discontinuity calculation means for calculating a speech speed which increases in accordance with the number of times each time the threshold time is counted.
The described speech speed conversion device. 5. An audio signal memory (7) for storing an input digital audio signal, and an audio analysis unit (11) for classifying the digital audio signal output from the audio signal memory into unvoiced sound, voiced sound, and non-voiced sound. 8) and thinning out only the voiced sounds classified by the voice analysis unit in accordance with the specified speech speed, and connecting the unvoiced sound, the decimated voiced sound and the unvoiced sound to form a new voice signal. And a signal synthesizing unit (9) for synthesizing the audio signal. The audio analysis unit measures the duration of the silence from the silence start time for each silence in the audio signal. Every time the silence duration timer counts the threshold time, it changes from the start time of the unvoiced sound or voiced sound that follows the corresponding silence to the target speech speed higher than the normal speech speed with the lapse of time, and thereafter The goal Speech speed conversion apparatus characterized by comprising a speech rate calculation unit sequentially calculates the speaking speed to maintain a fast delivery to the signal combining unit.