JP2010212783A - Speaking speed conversion device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speaking speed conversion device that suitably performs speaking speed conversion of a received speech even when a conversation is too short to measure the speaking speed of a transmission-side user or even when a transmitted speech instantaneously changes in speaking speed. <P>SOLUTION: Conversion processing for making the speaking speed of the received speech slow is performed using a parameter representing a stable basic speaking speed that the user originally has and a parameter representing the speaking speed of the received speech. Further, it is estimated whether the user is speaking in a hurry in the current telephone call using the parameter representing the stable basic speaking speed that the user originally has, and when the user is not in a hurry, conversion processing for making the speaking speed of the received speech slow is carried out. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a speech speed conversion device that converts a speech speed of an information terminal or the like having a voice call function.

  As a conventional method of converting the speech speed of an information terminal or the like having a voice call function, the conversation speech speed of the received voice and the conversation speed of the transmitted voice are each measured, and the difference between the respective conversation speeds is There is a method of adjusting the output speed of the received voice so as to reduce the difference in the conversation speed between the received voice and the transmitted voice when the value exceeds a certain value (see, for example, Patent Document 1). In this method, the conversation speed between the received voice and the transmitted voice is calculated using the result of voice recognition.

JP 2008-31754 A

  In the invention described in Patent Document 1, the conversation speed of the transmitted voice is measured during a call. However, there is a problem that even if a short utterance is measured, the speaking speed is not reliable, and in order to increase the reliability, the user on the transmitting side must speak a certain amount of utterances. For example, a user who has received an incoming call rarely speaks the amount of utterance immediately after the start of the call, and it is often considered that the user speaks the amount of utterance after a while after the start of the call. For this reason, it is impossible to measure the reliable speech speed until the user who has received the incoming call speaks a collective amount of speech after the call starts.

  Moreover, since it is influenced by the instantaneous change in the speech speed of the transmitted voice, there is a problem that the speech speed conversion of the received voice is performed so as to fluctuate excessively and becomes difficult to hear.

  Therefore, the present invention is capable of performing stable speech speed conversion by controlling the speech speed of the received speech based on the user's original sense even in a case where the user on the transmitting side does not speak the total speech volume. To avoid adverse effects that are difficult to hear due to excessive fluctuations in the speech speed conversion of the received voice due to the instantaneous user's utterance activity and changes in the speech speed of the transmitted voice. An object of the present invention is to provide a speech speed conversion device capable of

  In order to achieve the above object, a speech speed conversion apparatus according to the present invention includes a speech speed conversion means, a reception activity parameter acquisition means for obtaining a parameter related to the activity level of the received voice, and a parameter related to the activity level of the transmitted voice. Based on the evaluation of the reception activity parameter and the long-term transmission activity parameter, the long-term transmission activity parameter acquisition unit for obtaining the parameter related to the long-term activity of the transmitted voice, A means for performing speech speed conversion on the basis of the selection result, and means for performing speech speed conversion by the speech speed conversion means on the received voice signal based on the selection result, And a parameter updating means for updating the long-term transmission activity parameter.

  According to the present invention, even if the user on the transmitting side does not speak the total amount of speech, the speech speed of the incoming call is controlled based on the user's original sense, and stable speech speed conversion is performed. It is possible to avoid the adverse effect of making it difficult to hear by changing the speech speed of the received voice due to excessive fluctuations due to the instantaneous user's utterance activity and the change in the voice speed of the transmitted voice. It is possible to provide a speech speed conversion device that can

The block diagram which shows the structure of the speech-speed converter which concerns on embodiment of this invention. The flowchart which shows operation | movement of the speech-speed conversion apparatus which concerns on embodiment of this invention. The flowchart which shows the speech speed conversion method using estimation of the user's rush state using a long-term parameter and a short-term parameter. The flowchart which shows a process in case the speech-speed converter which concerns on embodiment of this invention operate | moves using the estimation of a user's rush state using a long-term parameter and a short-term parameter.

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a block diagram showing a configuration of a speech speed conversion apparatus according to an embodiment of the present invention. In FIG. 1, a speech speed conversion device 10 is shown as an example used for speech speed conversion in a call in a terminal device having a call function, for example, a mobile phone. The speech speed conversion apparatus 10 of FIG. 1 includes a transmission side activity estimation unit 120, a reception side activity estimation unit 230, a speech speed conversion selection unit 270, and a speech speed conversion unit 280. In a call, a voice signal input from the transmission side is input to the speech speed conversion apparatus 10 as a transmission voice signal 100 and input to the encoding unit 110 and the transmission side activity level estimation unit 120. The encoding unit 110 receives the transmission voice signal 100 and performs voice encoding, and transmits the encoded voice data to the terminal of the other party using a communication function (not shown).

  On the other hand, the decoding unit 210 generates the received voice signal 200 by decoding the encoded voice data received from the other party using the communication function. This reception voice signal 200 is output as a reception output voice signal 290 via the speech speed conversion function of the speech speed conversion unit 280.

  First, the transmission side will be described. In FIG. 1, a transmission voice signal 100 is input to an encoding unit 110 and a transmission side activity estimation unit 120. Encoding section 110 performs voice encoding with transmission voice signal 100 as an input, and transmits the encoded voice data to the other party's terminal.

  The transmitting side activity estimation unit 120 acquires parameters related to short-term activity related to utterances in the call of the transmitted voice signal 100 input after the call starts, and a user state to be described later. Used to select speech speed conversion. Here, an example in which the transmission side activity estimation unit 120 includes a transmission activity acquisition unit 130, a long-term activity update unit 140, a long-term activity acquisition unit 150, and a user state estimation unit 160 will be described.

The transmission activity level acquisition unit 130 analyzes the input transmission voice signal 100 to acquire parameters related to the transmission activity level, and sends them to the speech rate conversion selection unit 270 and the long-term activity level update unit 140. Output. Here, parameter information (P) representing the outline of the speech spectrum is used as a parameter related to the transmission activity. As a typical example of parameter information P that represents the outline of the speech spectrum, LSF (Line Spectral Frequency) parameters can be used, but there are other interpolation characteristics in the time direction of parameters such as LSP parameters and ISF parameters. Any good parameter can be used.

ここで、Ｐｉ（ｍ）は、第ｍフレームにおけるｉ番目のパラメータＰを表す。ＪはパラメータＰの次数を表す。また、Ｇｍは第ｍフレームについての重み、Ｗはパラメータ毎の重みを表し、Ｗｉはｉ番目のパラメータの重みを表す。Ｋは典型的には、K=１、または１〜２の値を用いることができるが、これに限定されるものではない。 The transmission activity level acquisition unit 130 uses parameters related to the transmission activity level to obtain a time variation amount of the parameter weighted for each parameter. Here, the time variation of the parameter weighted for each parameter is referred to as an activity parameter A. An example of a method for obtaining the activity parameter A will be described.

Here, Pi (m) represents the i-th parameter P in the m-th frame. J represents the order of the parameter P. Gm represents the weight for the mth frame, W represents the weight for each parameter, and Wi represents the weight of the i-th parameter. K may typically be K = 1, or a value of 1 to 2, but is not limited thereto.

Gm uses a weight of 0 or a small value when it is not a voice frame, and uses a large weight when it is a voice frame. As another method, Gm can be set using a prediction gain for the m-th frame, its partial frame, or its neighboring frames. The prediction gain is obtained by performing prediction analysis such as LPC prediction analysis on the frame. As an example of the prediction gain, it is possible to use a ratio between the speech signal energy before prediction and the prediction residual signal energy, or a value based on a logarithmic value of the ratio or a decibel value of the ratio. It is known that the prediction gain of the speech section is larger than the prediction gain of the silent section. Furthermore, even in the speech section, the prediction gain of the speech section with high vowel property is higher than the prediction gain of the speech section with high consonant nature. Tend to be large.

As an example, Gm can be set by quantizing or mapping the prediction gain of the section related to the m-th frame by a predetermined method, such as Gm = Q (prediction gain). Here, Q () represents quantization or mapping for the value in parentheses. For example, using a predetermined threshold Th1, Th2 (Th1 <Th2), Gm = g1 when the prediction gain <Th1, Gm = g2 when Th1 ≦ the prediction gain <Th2, and Gm when Th2 <the prediction gain. = G3. Here, there is a method of satisfying g1 <g2 <g3 or g1 <g2 ≦ g3. By appropriately setting the thresholds Th1 and Th2, it can be divided into three categories: a frame with a low prediction gain value, a frame with a medium prediction gain, and a frame with a high prediction gain. For example, it is possible to correspond to a silent section, a speech section with high consonant property, and a speech section with high vowel property, respectively. Also, Gm associated with the three categories can be obtained by appropriately setting the values of g1, g2, and g3. To determine the activity by excluding frames with low prediction gain values, Gm should be positive for frames with high prediction gain, and Gm for frames with low prediction gain. Set to 0 or a negative value.

However, the threshold setting is not limited to three stages. The weight Gm may be any other method as long as it is a weight Gm for preventing the activity of a section unrelated to the voice for which the activity is obtained from being positively reflected.

Further, when the activity parameter A is calculated, by using the weight Wi for each parameter, it is possible to obtain the time variation amount of the weighted parameter that places importance on the effective parameter based on the activity. For example, the LSF parameter is typically expressed as (q1, q2, q3,..., Q10) and about 10 parameters on the frequency axis for a voice signal in a telephone band. The outline of the spectrum can be expressed as follows, but the subjective influence of each parameter for expressing the outline of the spectrum is not uniform. That is, the higher-order parameters such as q8, q9, and q10 have a small subjective influence, the lower-order parameters have a slightly larger subjective influence, and the intermediate-order parameters have a significantly larger subjective influence. There is a feature. By reflecting the importance for each parameter in the weight, there is an effect that the activity parameter A can be obtained with higher accuracy. It is also possible to set the weight Wi of the higher-order parameter that has a small subjective influence to zero. However, this is only an example, and any method that reduces the number of parameters to several parameters that have a large subjective influence is included in the present invention. By doing so, the order of the parameters can be substantially reduced, so that the calculation amount can be greatly reduced without affecting the accuracy calculation.

Since parameters such as LSP parameters and ISF parameters have the same characteristics as LSF, the same effects can be obtained for weighting of each parameter and reduction of calculation amount.

  Needless to say, the activity parameter is not limited to the above-described method. In short, if it is information that can be used to estimate the degree of activity related to temporal changes in acoustic characteristics related to the amount of speech utterance and the amount of phonological change, use effective means to obtain that information. Thus, it can be utilized as an embodiment of the present invention.

  The information used to acquire the transmission activity level may be a signal derived from the transmission voice, and the same effect can be obtained even in that case. In short, the same effect can be obtained as long as information relating to the activity level of the user on the transmitting side can be obtained.

  The long-term activity acquisition unit 150 stores a long-term speech activity parameter obtained based on the transmitted voice of a past call in a memory, and stores this in a long-term activity update unit 140, a user state estimation unit 160, To the speech speed conversion selection unit 270. Here, the long-term activity update unit 140, which will be described later, will be described as a configuration in which a long-term transmission activity parameter obtained through a call of a past call is stored in a memory in the long-term activity acquisition unit 150. However, any configuration that can obtain the long-term transmission activity parameter based on the transmitted voice of the past call is included in the gist of the present invention. Needless to say.

The long-term activity update unit 140 uses the activity parameter (transmission activity parameter) output from the transmission activity income unit 130 and the long-term activity activity parameter from the long-term activity acquisition unit 150, and Update the transmission activity parameter. An example of an update method is to first use the long-term transmission activity parameter obtained through the call of the past call at the start of the call of the call.
LA = LA_past
And LA_past is a long-term transmission activity parameter obtained through a call of a past call stored in the long-term activity acquisition unit 150, and LA represents a long-term transmission activity parameter that is appropriately updated under a predetermined condition within the call. . Then, the following update is performed when a predetermined update condition as described below is satisfied.

LA = αSA + (1-α) LA
LA_past is
LA_past = LA *
Update as.

Here, SA represents a transmission activity parameter. LA * is an LA value to be used for updating the long-term transmission activity parameter LA_past among the updated values of LA obtained until the end of the call. For example, LA * is the average value of LA obtained during the current call, the median value of LA, the LA value at a predetermined time before the end of the call, or obtained during the current call. The average value of SA and the median value of SA can be used. LA_past is updated after the call ends or when LA * is determined during the call.

  α is a coefficient for adjusting the update rate of LA, and 0 <α <1. As a typical example, a value of about 0.1 can be used, but when the reliability of the SA value is high, α is 1 A value close to can be used. Since the value of α also depends on the frame length, it is not limited to the value of the above typical example. On the other hand, when the reliability of the acquired activity is low, such as when the speech is short and the input is performed in an environment with a lot of background noise included in the speech, α is set to 0 or update is performed. Do not do it. By doing this, updating the long-term transmission activity parameter becomes highly reliable, and as a result, it is possible to obtain a parameter that more accurately reflects the activity of the speaker in the normal way of speaking. effective. In addition, when the user state estimation unit 160 described later estimates that the user state is a state in which the activity is deviating from the normal way of speaking in the past, the long-term transmission activity parameter is not updated. Therefore, α is set to 0, or the long-term transmission activity parameter is not updated. By doing so, the long-term activity update unit 140 updates the long-term transmission activity parameter with high reliability. As a result, the parameter that reflects the activity when the speaker originally speaks normally is updated. There is an effect that acquisition is possible.

  The user state estimation unit 160 estimates the user state in the current call using the transmission activity parameter from the transmission activity income unit 130 and the long-term transmission activity parameter from the long-term activity acquisition unit 150. . Here, in particular, it is estimated whether the user in the current call is in a normal speaking user state or a rushed user state. The estimation result is output to the speech speed conversion unit 280 and the user state notification unit 161.

  The user state estimation unit 160 compares the short-term information acquired in the current call with the long-term information acquired in the past call (information reflecting the original user's call state), thereby It is estimated whether or not the user is in a hurry as the user state in a call. This distinguishes a voice signal that is originally spoken by a person who speaks in a normal manner and a voice signal that is spoken by a person who is originally not in a hurry.

In addition, as short-term information acquired during the current call, the short-term activity parameter (transmission activity parameter when sending speech is targeted) is further sub-short-term, with fluctuations stabilized in the near-short term. Use of the activity parameter has an effect that the user state can be estimated more stably. As a method for obtaining the quasi-short-term activity parameter, for example, when the quasi-short-term activity parameter with respect to the SA (short-term) activity parameter is SSA, a coefficient β (having a relationship of α <β <1) is used.
SSA = βSA + (1-β) SSA
It can ask for. Here, β is a coefficient for adjusting the SSA update rate.

  Here, as an example of using user activity parameters as long-term information and short-term information, the gist of the invention is a parameter (long-term parameter) reflecting the original state of the speaker S, If the same speaker S is a parameter (short-term parameter) obtained from the speech when speaking in the current call, the speaker in the current call of the speaker S is compared by comparing the long-term and short-term parameters. It is possible to estimate the state of S.

  For this reason, based on the above principle, the user (speaker) state is estimated by using the same method for long-term information and short-term information, even for speech speed information, and the user is in a hurry It is possible to estimate whether or not. That is, since the long-term parameter related to the user's speaking speed in the call in the past call is based on the user's original way of speaking, the short-term of the user acquired in the call of the current call is based on this parameter. By comparing with the parameter related to the speech speed, it can be determined whether the speech speed of the user is significantly faster than the long-term speech speed in the original state in the current call. As a result, it is possible to estimate whether the user is in a hurry in the current call.

  When a parameter related to activity is used as a parameter, if the activity represented by the short-term parameter is significantly higher than the activity represented by the long-term parameter, the speaker S speaks in a state where the activity is higher than usual. Therefore, it can be estimated that the speaker S is in a hurry. In this embodiment, an example is shown in which the user state is estimated for the user on the transmission side, but the present invention is not limited to this, and it goes without saying that this method can also be used for the received voice. In short, if parameters of the same nature are used, comparisons and evaluations using long-term and short-term parameters will detect that the speaker is speaking in a state that deviates from the normal state in the short term. If it is possible and is in line with the gist, it is included in the present invention.

  When the user state estimation unit 160 estimates that the user is in a hurry state, the speech speed conversion unit 280 prevents the function for reducing the speech speed from being used in the speech speed conversion. In this way, when the user is in a hurry, the user can be prevented from being stressed by the received voice being played back slowly with the speech speed conversion function that slows down the speech speed.

  The user state notification unit 161 has a function of notifying the user state estimated by the user state estimation unit 160 to the outside of the apparatus by characters, images, moving images, color display, sound, vibration, and the like. When the speech rate conversion device of the present invention is incorporated into, for example, a mobile phone, it can be sensed by the user of the mobile phone through information transmission means such as incoming notification information in addition to visual, auditory, tactile, taste, and olfactory senses. Notify in such a form. By doing so, there is an effect that it is possible to confirm in what state the user is talking during or after the call.

Next, the receiving side will be described.
The reception activity level estimation unit 230 inputs and analyzes the reception voice signal 200 decoded by the decoding unit 210, so that the reception side voice signal input from the start of the call is about the utterance in the call. A parameter related to short-term activity is acquired, and this is output to the speech speed conversion selection unit 270 and also to the reception state notification unit 231. Here, as a parameter related to the reception activity level, it is determined by a method similar to that on the transmission side, and is referred to as a reception activity level parameter B.

  The reception state notification unit 231 has a function of notifying the user state estimated by the reception activity estimation unit 230 to the outside of the apparatus by characters, images, moving images, color displays, sounds, vibrations, and the like. When the speech rate conversion device of the present invention is incorporated into, for example, a mobile phone, it can be sensed by the user of the mobile phone through information transmission means such as visual, auditory, tactile, gustatory, olfactory, and incoming notification information. Notify in form. By doing so, there is an effect that it is possible to check at what level of activity the other party is talking at the time of the call or after the call ends.

The speech rate conversion selection unit 270 inputs the reception activity parameter B from the reception side activity estimation unit 230 and the long-term transmission activity parameter LA_past (LA may be used in some cases) from the long-term activity acquisition unit 150. Then, based on the evaluation of the reception activity level parameter and the long-term transmission activity level parameter, selection regarding the execution of the speech rate conversion process in the speech rate conversion unit 280 is performed. That is, the speech activity conversion process is performed so that the speech speed is reduced when the activity level of the received speech is higher than the long-term speech activity of the transmitted speech by comparing the received speech activity parameter and the long-term speech activity parameter. (Control information related to the speech rate conversion rate and speech rate) is selected. At this time, in the case of the configuration using the user state estimation unit 160, the estimation result of the user state is also taken into consideration. That is, when it is estimated that the user is in a hurry state, even when it is determined that the activity level of the received call is higher than the long-term activity level of transmitting the speech, Avoid using the function to slow down.
As an example of a method for preventing the use of the function for slowing down the speech speed, when it is estimated that the user is in a hurry state, the speech speed conversion may not be performed on the received voice. It is valid. Also, when playing back the recorded call contents (voice data) during a call after the call is finished, for example, when it is estimated that the user is in a hurry state, the speed of the received voice is increased. A method of reproducing and a method of reproducing by shortening the silent section with respect to the received voice are also effective.

In this way, when the user is in a hurry, the user can be prevented from being stressed by the received voice being played back slowly with the speech speed conversion function that slows down the speech speed.

  The speech speed conversion unit 280 converts the speech speed of the received speech signal using the speech speed conversion rate selected based on the selection information from the speech speed conversion selection unit 270 or control information related to the speech speed. And the received output audio signal 290 is output. At this time, one of the options of the speech speed conversion unit 280 is not to perform the speech speed conversion.

  FIG. 2 is a flowchart showing the operation of the speech speed conversion apparatus 10. The speech speed converting apparatus 10 first acquires the long-term transmission activity parameter obtained based on the transmitted voice of the past call in step S20, and in step S22, the decoding unit 210 in the current call decodes it. The parameter related to the reception activity level is acquired from the received reception voice signal 200. Similarly, in step S23, parameters related to transmission activity are acquired from the transmission voice signal 100 in the current call, and in step S24, long-term transmission activity parameters and long-term transmission activity parameters are used. Update the transmission activity parameter. Note that the order of the processing on the transmitting side in steps S23 and S24 and the processing on the receiving side in step S22 may be performed first. Next, in step S25, selection information relating to the implementation of the speech rate conversion process is generated based on the parameter related to the reception activity level and the evaluation using the long-term transmission activity level parameter. In step S26, speech speed conversion is performed on the received speech signal using the speech speed conversion rate selected based on the selection information from step S25 or control information related to the speech speed, and the result is received. The output audio signal 290 is output. In this case, not performing the speech speed conversion is one of the selection results of the speech speed conversion. If the call has not ended in step S27, the above processing is repeated for the next time interval. When the call ends, the long-term transmission activity information is retained in step S28 so that it can be used in step S20 during the next call.

  Next, a speech speed conversion method using estimation of the user's rush state using the long-term parameter and the short-term parameter will be described using the flowchart of FIG. In step S30, a long-term transmission activity parameter for the speaker and an activity (short-term activity) for the same speaker in a predetermined section obtained in the latest call are acquired. In step S31, it is estimated whether or not the target speaker (user) is in a hurry state by evaluating information related to the long-term activity and the short-term activity. If it is determined in step S32 that the state is not in a rush state, in step S34, conversion is performed to reduce the speech speed for the received voice. If it is determined in step S32 that the state is in a hurry state, in step S33, conversion for reducing the speech speed is not performed on the received voice.

  The flowchart of FIG. 4 is a flowchart showing the operation of the speech speed conversion apparatus 10 using the estimation of the user's rush state using the long-term parameter and the short-term parameter. In step S40, a long-term transmission activity parameter (long-term parameter) obtained based on the transmitted voice of the past call is acquired. In step S42, the received voice signal decoded by the decoding unit 210 in the current call. From 200, a parameter related to the reception activity level is acquired. Similarly, in step S43, a parameter (short-term parameter) related to transmission activity is acquired from the transmission voice signal in the current call, and in step S44, the transmission activity parameter and long-term transmission activity parameter are obtained. Use to update the long-term transmission activity parameter. Note that the order of the processing on the transmitting side in steps S43 and S44 and the processing on the receiving side in step S42 may be performed first.

  Next, in step S49a, it is estimated whether the target user is in a rush state by evaluating information related to the long-term activity and the short-term activity. If it is determined in step S49b that the state is in a hurry state, in step S46, conversion for reducing the speech speed is not performed on the received voice. If it is determined in step S49b that it is not in a rush state, in step S45, selection information related to the implementation of the speech rate conversion process based on the parameter related to the reception activity level and the evaluation using the long-term transmission activity parameter. Is generated. In step S46, speech rate conversion is performed on the received speech signal using the speech rate conversion rate selected based on the selection information from step S45 or control information related to the speech rate, and the result is received. Output as output audio signal. In this case, not performing the speech speed conversion is one of the selection results of the speech speed conversion. If the call has not ended in step S47, the above processing is repeated for the next time interval. If the call ends, the long-term activity information is retained in step S48 so that it can be used in step S40 during the next call.

  By adopting the configuration of such an embodiment, since the speech speed of the incoming call is controlled based on the user's original stable basic speech speed (related to long-term activity), stable speech speed conversion is performed. Can do. In addition, it is possible to avoid the adverse effect that the speech speed conversion of the received speech is excessively fluctuated due to the instantaneous user speech activity and the change in speech speed, and is difficult to hear.

  In addition, it is not limited to the said embodiment, You may change suitably in the range which does not deviate from the summary of this invention.

DESCRIPTION OF SYMBOLS 10 Speech speed converter, 110 encoding part, 120 transmission side activity estimation part, 130 transmission activity acquisition part, 140 long-term activity update part, 150 long-term activity acquisition part, 160 user state estimation part, 161 user State notifying unit, 210 encoding unit, 230 receiving side activity estimating unit, 231 receiving state notifying unit, 270 speaking rate conversion selecting unit, 280 speaking rate converting unit

Claims (5)

Speaking speed conversion means,
A reception activity parameter acquisition means for obtaining a parameter related to the activity of the received voice;
Transmission activity parameter acquisition means for obtaining a parameter related to the activity of the transmitted voice;
Long-term transmission activity parameter acquisition means for obtaining a parameter related to the long-term activity of the transmitted voice;
Based on the evaluation of the reception activity parameter and the long-term transmission activity parameter, a selection is made regarding the implementation of the speech rate conversion, and the speech rate conversion means performs a speech on the received speech signal based on the selection result. Means for speed conversion;
Parameter updating means for updating the long-term speech activity parameter based on the previous speech activity parameter;
A speech rate conversion device characterized by comprising: Speaking speed conversion means,
Transmission activity parameter acquisition means for obtaining a parameter related to the activity of the transmitted voice;
Long-term transmission activity parameter acquisition means for obtaining a parameter related to the long-term activity of the transmitted voice;
Means for performing speech speed conversion by the speech speed conversion means on the received voice signal;
State estimation means for estimating the rush situation of the user on the transmission side using the long-term transmission activity parameter and the transmission activity parameter;
When it is estimated that the transmitting user is not in a rushing state, it is possible to select a speech speed conversion that slows down the received voice. A speech speed conversion device comprising speech speed selection means for preventing selection. Speaking speed conversion means,
Short-term reception parameter obtaining means for obtaining short-term parameters related to the speech speed of the received speech;
Short-term transmission parameter acquisition means for obtaining short-term parameters related to the speech speed of the transmitted voice;
Long-term transmission parameter acquisition means for obtaining long-term parameters related to the speech speed of the transmitted voice;
Based on the evaluation of the short-term reception parameter and the long-term transmission parameter, the selection relating to the implementation of the speech rate conversion is performed, and the speech rate conversion by the speech rate conversion unit is performed on the received speech signal based on the selection result. Means to do,
Parameter updating means for updating the long-term transmission parameter;
A speech rate conversion device characterized by comprising: A short-term transmission parameter obtaining means for obtaining a short-term parameter relating to the speech speed of the speech speed conversion means and the speech speed;
Long-term transmission parameter acquisition means for obtaining long-term parameters related to the speech speed of the transmitted voice;
Means for performing speech speed conversion by the speech speed conversion means on the received voice signal;
A state estimating means for estimating a rush situation of the user on the transmission side using the long-term parameter and the short-term parameter;
When it is estimated that the transmitting user is not in a rushing state, it is possible to select a speech speed conversion that slows down the received voice. A speech speed conversion device comprising speech speed selection means for preventing selection. A speech speed conversion means capable of selecting the presence or absence of processing for slowing the speech speed for the received speech;
User state estimating means for estimating whether the transmitting user is in a hurry state;
When it is estimated that the sending user is not in a rush state, it is possible to select a speech speed conversion that slows down the received voice,
A speech speed conversion device comprising speech speed conversion selection means for not selecting a speech speed conversion that slows down a received voice when it is estimated that a transmitting user is in a rush state.

Images