JP6819426B2 - Speech processing program, speech processing method and speech processor - Google Patents

Description

The present invention relates to a voice processing program and the like.

In recent years, employees in a company have been making telephone calls and conference calls using application software of their own personal computer (Personal Computer) and a headset. In the following explanation, employees and other users in the company are collectively referred to as users.

If the user is not accustomed to handling the headset, the distance between the user's mouth and the headset microphone is often inadequate. For example, if the distance between the user's mouth and the microphone is short, the volume tends to exceed an appropriate level, which may cause discomfort to the other party. On the other hand, if the distance between the user's mouth and the microphone is long, the volume is not sufficient and it becomes difficult for the other party to hear the voice.

As a technique for evaluating sound quality and notifying the user, for example, there is a conventional technique 1. In the prior art 1, the sound quality is evaluated based on external factors such as the sound of paper rubbing against the microphone, echo, ambient noise, and residual noise, and the evaluation result is displayed to the user.

Japanese Unexamined Patent Publication No. 1-155430 JP-A-2010-259691

However, in the above-mentioned conventional technique, there is a problem that the utterance state of the input voice cannot be estimated appropriately.

For example, the volume of the input voice input from the user to the microphone is not always constant and fluctuates due to the influence of the user's psychological condition such as stress, so the appropriate distance between the user's mouth and the microphone. Is not constant. Therefore, it is desirable to appropriately estimate the utterance state of the input voice and notify the user so that the distance between the user's mouth and the microphone is appropriate.

On the other hand, in the evaluation of sound quality by the prior art 1, the sound quality is only evaluated in consideration of external factors such as noise, and the utterance state of the input voice is not evaluated. There is also a technology to notify whether the distance between the mouth and the microphone is appropriate based on the volume at the start of conversation, but as mentioned above, the volume of the input voice fluctuates due to the influence of the psychological situation. , The distance between the mouth and the microphone based on the volume at the start does not always continue and cannot be said to be the optimum distance.

In one aspect, it is an object of the present invention to provide a speech processing program, a speech processing method, and a speech processing device capable of appropriately estimating the vocalization state of an input speech.

In the first plan, the computer is made to perform the following processing. The computer extracts the pitch frequency and frequency power from the input voice. The computer outputs a determination result as to whether or not the condition that the value based on the pitch frequency and the frequency power is equal to or higher than a predetermined threshold value is satisfied. The computer estimates the utterance state of the input voice based on the relationship between the determination result and the average power of the frequency power.

The vocalization state of the input voice can be estimated appropriately.

FIG. 1 is a functional block diagram showing a configuration of a voice processing device according to the first embodiment. FIG. 2 is a flowchart showing a processing procedure of the voice processing device according to the first embodiment. FIG. 3 is a functional block diagram showing the configuration of the voice processing device according to the second embodiment. FIG. 4 is a diagram showing an example of the data structure of the estimation result. FIG. 5 is a flowchart (1) showing an example of the update process of the update unit according to the second embodiment. FIG. 6 is a flowchart (2) showing an example of the update process of the update unit according to the second embodiment. FIG. 7 is a flowchart (3) showing an example of the update process of the update unit according to the second embodiment. FIG. 8 is a flowchart showing a processing procedure of the voice processing device according to the second embodiment. FIG. 9 is a diagram showing an example of the system according to the third embodiment. FIG. 10 is a functional block diagram showing the configuration of the voice processing device according to the third embodiment. FIG. 11A is a functional block diagram showing a server configuration according to the third embodiment. FIG. 11B is a diagram showing an example of the data structure of the threshold table according to the third embodiment. FIG. 12 is a flowchart (1) showing a processing procedure of the voice processing device according to the third embodiment. FIG. 13 is a flowchart (2) showing a processing procedure of the voice processing device according to the third embodiment. FIG. 14 is a flowchart (3) showing a processing procedure of the voice processing device according to the third embodiment. FIG. 15 is a diagram showing an example of the system according to the fourth embodiment. FIG. 16 is a functional block diagram showing the configuration of the voice processing device according to the fourth embodiment. FIG. 17 is a functional block diagram showing a server configuration according to the fourth embodiment. FIG. 18 is a diagram showing an example of the data structure of the classification table according to the fourth embodiment. FIG. 19 is a diagram showing an example of a statistical data structure. FIG. 20 is a flowchart showing a processing procedure of the voice processing device according to the fourth embodiment. FIG. 21 is a flowchart showing a processing procedure of the server according to the fourth embodiment. FIG. 22 is a functional block diagram showing a configuration of the voice processing device according to the fifth embodiment. FIG. 23 is a flowchart showing a processing procedure of the voice processing device according to the fifth embodiment. FIG. 24 is a diagram showing an example of a computer hardware configuration that realizes a function similar to that of a voice processing device.

Hereinafter, examples of the voice processing program, voice processing method, and voice processing apparatus disclosed in the present application will be described in detail with reference to the drawings. The present invention is not limited to this embodiment.

FIG. 1 is a functional block diagram showing a configuration of a voice processing device according to the first embodiment. As shown in FIG. 1, the voice processing device 100 is connected to the microphone 10. The voice processing device 100 includes an AD (Analog / Digital) conversion unit 110, a pitch extraction unit 120a, a power extraction unit 120b, a stress detection unit 130, a storage unit 140, an estimation unit 150, and an information presentation unit 160. The pitch extraction unit 120a and the power extraction unit 120b are examples of the extraction unit.

The microphone 10 is a microphone attached to a headset (not shown) worn by the user, and collects the user's voice. The microphone 10 outputs the collected voice data of the user to the AD conversion unit 110 of the voice processing device 100. In the following description, the voice data output by the microphone 10 to the AD conversion unit 110 is referred to as an input voice.

The AD conversion unit 110 receives the input voice from the microphone 10 and executes AD conversion on the received input voice. The AD conversion unit 110 outputs the AD-converted input voice to the pitch extraction unit 120a and the power extraction unit 120b. AD conversion is a process of converting an analog signal into a digital signal. That is, the AD conversion unit 110 converts the input voice of the analog signal into the input voice of the digital signal. In the following description, the input voice of the digital signal converted by the AD conversion unit 110 is simply referred to as “input voice”.

The pitch extraction unit 120a is a processing unit that extracts a pitch that is the fundamental frequency of the input voice based on the input voice. The pitch extraction unit 120a outputs the extracted pitch information to the stress detection unit 130.

The pitch extraction unit 120a executes frame processing and pitch calculation processing. First, frame processing will be described. When the pitch extraction unit 120a takes out the signal sequence of the input voice as a "frame" for each predetermined number of samples and multiplies the frame by an analysis window such as a Hanning window to perform time-frequency conversion described later. Suppresses distortion caused by high frequency components.

For example, the pitch extraction unit 120a extracts a sample N in a section of 32 ms at a sampling frequency of 8 kHz as a frame. For example, N = 256. Let each sample included in the frame be "s (0), s (1), s (2), ..., S (N-1)". The pitch extraction unit 120a multiplies each of the above samples by a humming window. For example, the humming window is represented by equation (1).

The sample obtained by multiplying each sample by the humming window is defined as "x (0), x (1), x (2), ..., X (N-1)". In the following description, the sample "x (0), x (1), x (2), ..., X (N-1)" obtained by multiplying the humming window is referred to as a sample value.

The pitch calculation process will be described. The pitch extraction unit 120a calculates the autocorrelation function using each sample value included in the frame. For example, the pitch extraction unit 120a calculates the autocorrelation function φ (m) based on the equation (2). The m shown in the formula (2) indicates the delay time.

Regarding the equation (2), the pitch extraction unit 120a specifies the value of the delay time m at which the autocorrelation function becomes the maximum value, except for the delay time m = 0. The delay time m at which the autocorrelation function is maximized is expressed as "delay time m'". The pitch extraction unit 120a calculates the pitch based on the equation (3) after calculating the delay time m'.

Pitch = 1 / delay time m'... (3)

The pitch extraction unit 120a extracts a plurality of frames from the input voice by repeatedly executing the frame process for the input voice, and calculates the pitch from each frame. The pitch extraction unit 120a outputs pitch information for each frame to the stress detection unit 130.

Further, the pitch extraction unit 120a determines whether or not the frame is a sound section based on the maximum value φ (m') of the autocorrelation function, and outputs the determination result to the stress detection unit 130. For example, when the maximum value φ (m') of the autocorrelation function of the frame is equal to or more than a predetermined value, the pitch extraction unit 120a determines that the corresponding frame is a sound section.

The power extraction unit 120b is a processing unit that extracts the power of the input voice based on the input voice. The power extraction unit 120b outputs the extracted power information to the stress detection unit 130.

The power extraction unit 120b extracts a frame from the input voice by executing the frame processing in the same manner as the pitch extraction unit 120a. The power extraction unit 120b uses time-frequency conversion to time the input voice of each sample value “x (0), x (1), x (2), ..., X (N-1)” of the frame. Converts a region to a spectral signal in the frequency domain. As the time-frequency transform, for example, a Fast Fourier Transform (FFT) can be used. Then, the power extraction unit 120b obtains the power for each frequency band by squaring the spectral signal P (n) of each frequency band, obtains the total power over the entire frequency band, and calculates a logarithmic value. This is hereafter referred to as "power". For example, the power extraction unit 120b calculates the power using the spectral signal of the frame based on the equation (4).

The power extraction unit 120b extracts a plurality of frames from the input voice by repeatedly executing the frame processing for the input voice, and calculates the power from each frame. The power extraction unit 120b outputs power information for each frame to the stress detection unit 130.

The stress detection unit 130 is a processing unit that detects the stress value of the user based on the pitch and power of the input voice. For example, the stress detection unit 130 increases the stress value as the current pitch and power statistics are farther from the user's normal pitch and power statistics, and decreases the stress value as the current pitch and power statistics are closer. .. The stress detection unit 130 outputs the detected stress value information and the power information to the estimation unit 150.

Here, an example of processing of the stress detection unit 130 will be described. The stress detection unit 130 calculates and holds in advance the standard deviation of the pitch in normal times and the standard deviation of power in normal times from the pitch and power based on the input voice of the user in normal times. For example, the standard deviation of the pitch in normal times is referred to as "standard deviation σA1", and the standard deviation of power in normal times is referred to as "standard deviation σB1".

In the stress detection phase, the stress detection unit 130 calculates the "standard deviation σA2" of the pitch of each frame and calculates the "standard deviation σB2" of the power of each frame. For example, the stress detection unit 130 calculates the stress value based on the equation (5). In the formula (5), α and β are coefficients preset by the user.

Stress value = α × | standard deviation σA1-standard deviation σA2 | + β × | standard deviation σB1-standard deviation σB2 | ... (5)

The stress detection unit 130 calculates the stress value for each frame by repeatedly executing the above processing each time information on the pitch and power of the frame is acquired from the pitch extraction unit 120a and the power extraction unit 120b. The stress detection unit 130 associates the stress value for each frame with the power and outputs the stress value to the estimation unit 150.

The storage unit 140 has determination reference data 140a. The storage unit 140 corresponds to semiconductor memory elements such as RAM (Random Access Memory), ROM (Read Only Memory), and flash memory (Flash Memory), and storage devices such as HDD (Hard Disk Drive).

The determination reference data 140a includes a plurality of threshold data used when the estimation unit 150, which will be described later, estimates the utterance state of the input voice. Specifically, the determination reference data 140a includes a first threshold value, a second threshold value, and a third threshold value. The first threshold and the second threshold are thresholds to be compared with power. The magnitude relationship between the first threshold value and the second threshold value is such that the first threshold value> the second threshold value. The third threshold is a threshold to be compared with the stress value. For example, when the power is equal to or more than the second threshold value and less than the first threshold value, it can be said that the voice of the conversation is good.

The estimation unit 150 is a processing unit that estimates the utterance state of the input voice based on the stress value of the input voice, the power, and the determination reference data 140a. After estimating the vocalization state of the input voice, the estimation unit 150 generates a message according to the vocalization state and outputs the message to the information presentation unit 160 to display the message. As will be described later, the estimation unit 150 estimates the magnitude of the stress of the user from the stress value. The estimation unit 150 estimates whether or not the user's mouth is close to the microphone 10 by the power.

The estimation unit 150 estimates that the user's stress is "large" when the stress value is equal to or higher than the third threshold value, and when the stress value is less than the third threshold value, the user's stress is high. Estimated to be "small".

The estimation unit 150 calculates the average value of the power of the frame in the sounded section. In the following description, the average value of the power of the frame in the sound section is referred to as "average power". When the average power is equal to or higher than the first threshold value, the estimation unit 150 estimates that "the distance between the user's mouth and the microphone 10 is short". When the average power is less than the second threshold value, the estimation unit 150 estimates that "the distance between the user's mouth and the microphone 10 is long".

When the stress changes from low stress to high stress in the process of continuing the conversation, the power of the input voice tends to be larger than the current power of the user. Therefore, if the current stress is "small" and "the distance between the user's mouth and the microphone 10 is short", the power of the input voice will be appropriate when the stress changes to "large" in the future. There is a risk of exceeding the power. That is, the estimation unit 150 generates the first message "Please move the microphone a little away from the mouth" when "the stress is small" and "the distance between the user's mouth and the microphone 10 is short".

When the stress changes from high stress to low stress in the process of continuing the conversation, the power of the input voice tends to be smaller than the current power of the user. Therefore, if the current stress is "large" and "the distance between the user's mouth and the microphone 10 is long", the power of the input voice will be appropriate when the stress changes to "small" in the future. There is a risk of falling below the power. That is, the estimation unit 150 generates the second message "Please bring the microphone a little closer to the mouth" when "the stress is great" and "the distance between the user's mouth and the microphone 10 is long".

The information presentation unit 160 is a processing unit that presents a message generated by the estimation unit 150 to the user. For example, the information presentation unit 160 is connected to a display device such as a liquid crystal display or an output device such as a speaker. Here, as an example, the information presentation unit 160 is connected to a liquid crystal display and displays a message generated by the estimation unit 150.

By the way, each process of the AD conversion unit 110, the pitch extraction unit 120a, the power extraction unit 120b, the stress detection unit 130, the estimation unit 150, and the information presentation unit 160 shown in FIG. 1 is executed by a predetermined control unit (not shown). Is also good. This control unit can be realized by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like. The control unit can also be realized by hard-wired logic such as ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array).

Next, the processing procedure of the voice processing device 100 according to the first embodiment will be described. FIG. 2 is a flowchart showing a processing procedure of the voice processing device according to the first embodiment. As shown in FIG. 2, the AD conversion unit 110 of the voice processing device 100 starts accepting input voice (step S101). The AD conversion unit 110 performs AD conversion (step S102). The pitch extraction unit 120a of the voice processing device 100 extracts the pitch, and the power extraction unit 120b of the voice processing device 100 extracts the power (step S103).

The pitch extraction unit 120a detects a sound section (step S104). The stress detection unit 130 of the voice processing device 100 accumulates pitch power (step S105). When the pitch power corresponding to the specified number of frames is accumulated (steps S106, Yes), the stress detection unit 130 proceeds to step S107. On the other hand, the stress detection unit 130 proceeds to step S101 when the pitch power corresponding to the specified number of frames is not accumulated (steps S106, No).

The stress detection unit 130 calculates the stress value (step S107). The estimation unit 150 of the voice processing device 100 calculates the average power of the sounded section (step S108). When the average power is equal to or higher than the first threshold value (step S109, Yes), the estimation unit 150 shifts to step S110. On the other hand, when the average power is less than the first threshold value (steps S109, No), the estimation unit 150 shifts to step S112.

The estimation unit 150 determines whether or not the stress value is equal to or higher than the third threshold value (step S110). When the stress value is equal to or higher than the third threshold value (step S110, Yes), the estimation unit 150 proceeds to step S101. When the stress value is less than the third threshold value (step S110, No), the estimation unit 150 causes the information presentation unit 160 to display the first message (step S111), and proceeds to step S101. For example, the first message is "Please move the microphone away from your mouth."

The process proceeds to the description of step S112. The estimation unit 150 determines whether or not the average power is less than the second threshold value (step S112). When the average power is not less than the second threshold value (step S112, No), the estimation unit 150 shifts to step S101. On the other hand, when the average power is less than the second threshold value (step S112, Yes), the estimation unit 150 shifts to step S113.

The estimation unit 150 determines whether or not the stress value is equal to or higher than the third threshold value (step S113). When the stress value is not equal to or higher than the third threshold value (steps S113, No), the estimation unit 150 proceeds to step S101. On the other hand, when the stress value is equal to or higher than the third threshold value (step S113, Yes), the estimation unit 150 causes the information presentation unit 160 to display the second message (step S114), and proceeds to step S101. For example, the second message is "Please bring the microphone a little closer to your mouth."

Next, the effect of the voice processing device 100 according to the first embodiment will be described. The voice processing device 100 calculates the stress value of the user based on the pitch and power of the input voice, and estimates the utterance state based on both the magnitude relation of the calculated stress value and the magnitude relation of the power. Thereby, for example, it is possible to estimate the future utterance state in consideration of not only the power of the input voice but also the degree of stress of the user. In addition, since it is possible to appropriately estimate the transition of the vocalization state not only at present but also in the future, by generating a message according to the estimation result and presenting it to the user, the distance between the user's mouth and the microphone can be estimated. It can be kept at an appropriate distance, and each user's call can be kept comfortable.

FIG. 3 is a functional block diagram showing the configuration of the voice processing device according to the second embodiment. The voice processing device 200 is connected to the microphone 10 as shown in FIG. The voice processing device 200 includes an AD conversion unit 110, a pitch extraction unit 120a, a power extraction unit 120b, a stress detection unit 130, a storage unit 140, an information presentation unit 160, an estimation unit 210, and an update unit 220. Of these, the description of the microphone 10, the AD conversion unit 110, the pitch extraction unit 120a, the power extraction unit 120b, the stress detection unit 130, the storage unit 140, and the information presentation unit 160 is the same as that described in the first embodiment. , The description is omitted.

The estimation unit 210 is a processing unit that estimates the utterance state of the input voice based on the stress value of the input voice, the average power, and the determination reference data 140a. The estimation unit 210 outputs the estimation result to the update unit 220 in addition to the processing of the estimation unit 150 described in the first embodiment.

FIG. 4 is a diagram showing an example of the data structure of the estimation result. As shown in FIG. 4, the estimation result includes a presentation flag, a message type, a stress value Sn, an average power Pn, a stress value Sp1, an average power Pp1, a stress value Sp2, and an average power Pp2.

The presentation flag is information indicating whether or not a message was displayed at the time of the previous estimation. If the previous message was presented, the presentation flag is "on", and if the previous message was not presented, the presentation flag is "off". The message type is information indicating whether the previously presented message is the first message or the second message. If the presentation flag is "off", no information is stored in the message type.

The stress value Sn indicates whether the stress value of the current input voice is “large” or “small”. The average power Pn indicates the average power of the current input voice. The stress value Sp1 indicates whether the stress value of the input voice when the previous message is presented is “large” or “small”. The average power Pp1 indicates the average power of the input voice when the previous message is presented.

The stress value Sp2 indicates whether the stress value of the input voice before a certain period of time is “large” or “small”. The average power Pp2 indicates the average power of the input voice before a certain time.

The update unit 220 is a processing unit that updates the first threshold value, the second threshold value, and the third threshold value of the determination reference data 140a based on the estimation result acquired from the estimation unit 210. The determination standard data 140a is an example of learning data. As described above, if the average power is equal to or more than the second threshold value and less than the first threshold value, the conversation is good. Further, the third threshold value is a threshold value used when distinguishing the magnitude of stress.

If there is no change in the stress value and the average power Pn is less than the first threshold value before and after displaying the first message "Please move the microphone a little away from your mouth", the power decreases to a good state. It can be said that the vocalization state is improved. In this case, the update unit 220 skips the update process assuming that the first threshold value, the second threshold value, and the third threshold value are correct values.

If there is no change in the stress value and no change in the average power (compared with Pn and Pp1) before and after the first message "Please move the microphone away from your mouth" is displayed, the vocalization state is improved. It has not been. In this case, the update unit 220 raises the first threshold value, assuming that the first threshold value is an incorrect value.

If the stress value changes from small to large and the average power (comparison with Pn and Pp1) increases before and after the first message "Please move the microphone away from your mouth" is displayed, the stress level is changed. I haven't been able to make a proper decision and present an appropriate message. In this case, the update unit 220 lowers the third threshold value, assuming that the third threshold value is an incorrect value.

Before and after displaying the second message "Please bring the microphone a little closer to your mouth", if there is no change in the stress value and the average power Pn becomes equal to or higher than the second threshold value, the power increases to a good state. It can be said that it is. In this case, the update unit 220 skips the update process assuming that the first threshold value, the second threshold value, and the third threshold value are correct values.

If there is no change in the stress value and no change in the average power (comparison with Pn and Pp1) before and after the second message "Please bring the microphone a little closer to your mouth" is displayed, the vocalization state is improved. It has not been. In this case, the update unit 220 lowers the second threshold value, assuming that the second threshold value is an incorrect value.

If the stress value changes from large to small and the average power (comparison with Pn and Pp1) decreases before and after the second message "Please bring the microphone a little closer to your mouth" is displayed, the magnitude of the stress is changed. I haven't been able to make a proper decision and present an appropriate message. In this case, the update unit 220 raises the third threshold value, assuming that the third threshold value is an incorrect value.

If the update unit 220 has not presented the previous message, there is no change in the stress value, and there is no change in the average power, it is assumed that the first threshold value, the second threshold value, and the third threshold value are correct values. Skip the update process.

If the previous message was not presented, the stress value did not change, and the average power (comparison with Pn and Pp2) increased, the second message "Please bring the microphone closer to your mouth" is not presented. Therefore, assuming that the second threshold value is incorrect, the second threshold value is raised.

If the previous message was not presented, the stress value did not change, and the average power (comparison with Pn and Pp2) decreased, the first message "Please move the microphone away from your mouth" is not presented. Therefore, assuming that the first threshold value is incorrect, the first threshold value is lowered.

By repeatedly executing the above processing, the update unit 220 updates the first threshold value, the second threshold value, and the third threshold value so that the first threshold value, the second threshold value, and the third threshold value become correct values.

5, FIG. 6 and FIG. 7 are flowcharts showing an example of the update process of the update unit according to the second embodiment. As shown in FIG. 5, the update unit 220 acquires the estimation result (step S10) and determines whether or not the previous message was presented (step S11). If the update unit 220 has not presented the previous message (steps S11, No), the update unit 220 proceeds to step S21 in FIG. On the other hand, if the update unit 220 has presented the previous message (steps S11, Yes), the update unit 220 proceeds to step S12.

The update unit 220 determines whether or not the presented message is the "first message" (step S12). If the presented message is not the "first message" (steps S12, No), the update unit 220 proceeds to step S17 of FIG. If the presented message is the "first message" (steps S12, Yes), the update unit 220 proceeds to step S13.

When the stress values Sn and Sp1 are low stress and the average power Pn and Pp1 are unchanged (steps S13, Yes), the update unit 220 raises the first threshold value (step S14). For example, in step 14, the update unit 220 updates the first threshold value based on the equation (6).

First threshold = 1.05 x first threshold ... (6)

On the other hand, when the stress values Sn and Sp1 are not small stress or the average power Pn and Pp1 are changed (steps S13 and No), the update unit 220 shifts to step S15.

When the stress value Sn is high stress and the stress value Sp1 is low stress (steps S15, Yes), the update unit 220 lowers the third threshold value (step S16). For example, in step S16, the update unit 220 updates the third threshold value based on the equation (7).

Third threshold = 0.9 x third threshold + 0.1 x (Sp1-Sn) ... (7)

On the other hand, when the stress value Sn is high stress and the stress value Sp1 is not low stress (steps S15, No), the update unit 220 ends the process.

The description shifts to FIG. When the stress values Sn and Sp1 are stressful and the average powers Pn and Pp1 are unchanged (steps S17, Yes), the update unit 220 lowers the second threshold value (step S18). For example, in step S18, the update unit 220 updates the second threshold value based on the equation (8).

Second threshold = 0.95 x second threshold ... (8)

On the other hand, when the stress values Sn and Sp1 are not stressful or the average power Pn and Pp1 are changed (steps S17 and No), the update unit 220 proceeds to step S19.

When the stress value Sn is low stress and the stress value Sp1 is high stress (step S19, Yes), the update unit 220 raises the third threshold value (step S20). For example, in step S20, the update unit 220 updates the third threshold value based on the equation (7). In step S19, when the stress value Sn is low stress and the stress value Sp1 is not high stress (steps S19, No), the update unit 220 ends the process.

The description shifts to FIG. When the stress values Sn and Sp2 are high in stress and the average power Pn is increased as compared with Pp2 (steps S21, Yes), the update unit 220 raises the second threshold value (step S22). For example, the update unit 220 updates the second threshold value in step S22 based on the equation (9).

Second threshold = 0.9 × second threshold + 0.1 × (Pn-Pp2) ... (9)

On the other hand, when the stress values Sn and Sp2 are not stressful or the average power Pn is not increased as compared with Pp2 (steps S21, No), the update unit 220 shifts to step S23.

When the stress values Sn and Sp2 are small and the average power Pn is reduced as compared with Pp2 (steps S23, Yes), the update unit 220 lowers the first threshold value (step S24). For example, in step S24, the update unit 220 updates the first threshold value based on the equation (10). When the stress values Sn and Sp2 are not low in stress or the average power Pn is not decreased as compared with Pp2 (steps S23, No), the update unit 220 ends the process.

First threshold = 0.95 x first threshold ... (10)

Next, the processing procedure of the voice processing device 200 according to the second embodiment will be described. FIG. 8 is a flowchart showing a processing procedure of the voice processing device according to the second embodiment. As shown in FIG. 8, the AD conversion unit 110 of the voice processing device 200 starts accepting the input voice (step S201). The AD conversion unit 110 performs AD conversion (step S202). The pitch extraction unit 120a of the voice processing device 200 extracts the pitch, and the power extraction unit 120b of the voice processing device 200 extracts the power (step S203).

The pitch extraction unit 120a detects a sound section (step S204). The stress detection unit 130 of the voice processing device 200 accumulates pitch power (step S205). When the pitch power corresponding to the specified number of frames is accumulated (step S206, Yes), the stress detection unit 130 proceeds to step S207. On the other hand, the stress detection unit 130 proceeds to step S201 when the pitch power corresponding to the specified number of frames is not accumulated (steps S206, No).

The stress detection unit 130 calculates the stress value (step S207). The estimation unit 210 of the voice processing device 200 calculates the average power of the sounded section (step S208). When the average power is equal to or higher than the first threshold value (step S209, Yes), the estimation unit 210 shifts to step S210. On the other hand, when the average power is less than the first threshold value (step S209, No), the estimation unit 210 shifts to step S212.

The estimation unit 210 determines whether or not the stress value is equal to or higher than the third threshold value (step S210). When the stress value is equal to or higher than the third threshold value (step S210, Yes), the estimation unit 210 shifts to step S215. When the stress value is less than the third threshold value (step S210, No), the estimation unit 210 causes the information presentation unit 160 to display the first message (step S211), and proceeds to step S215. For example, the first message is "Please move the microphone away from your mouth."

The process proceeds to the description of step S212. The estimation unit 210 determines whether or not the average power is less than the second threshold value (step S212). When the average power is not less than the second threshold value (step S212, No), the estimation unit 210 shifts to step S215. On the other hand, when the average power is less than the second threshold value (step S212, Yes), the estimation unit 210 shifts to step S213.

The estimation unit 210 determines whether or not the stress value is equal to or higher than the third threshold value (step S213). When the stress value is not equal to or higher than the third threshold value (steps S213 and No), the estimation unit 210 proceeds to step S215. On the other hand, when the stress value is equal to or higher than the third threshold value (step S213, Yes), the estimation unit 210 causes the information presentation unit 160 to display the second message (step S214), and proceeds to step S215. For example, the second message is "Please bring the microphone a little closer to your mouth."

The update unit 220 of the voice processing device 200 executes the update process (step S215). The update process shown in step S215 corresponds to the process shown in FIGS. 5, 6, and 7. The estimation unit 150 stores the stress value and the average power in the storage unit 140 (step S216), and proceeds to step S201.

Next, the effect of the voice processing device 200 according to the second embodiment will be described. The voice processing device 200 updates the first threshold value, the second threshold value, and the third threshold value by repeatedly executing the update process so that the first threshold value, the second threshold value, and the third threshold value become correct values. .. As a result, it is possible to appropriately determine whether or not the stress value is high or low and the power is good, and it is possible to display an appropriate message for the current vocalization state.

FIG. 9 is a diagram showing an example of the system according to the third embodiment. As shown in FIG. 9, this system has voice processing devices 300a, 300b, 300c and a server 400. The voice processing devices 300a to 300c and the server 400 are connected to each other via the network 50. Here, as an example, the voice processing devices 300a to 300c are shown, but other voice processing devices may be included. In the following description, the voice processing devices 300a to 300c are collectively referred to as the voice processing device 300.

The voice processing device 300 repeatedly executes the update process in the same manner as the voice processing device 200 described in the second embodiment, and transmits the updated determination reference data 140a to the server 400. As described in Examples 1 and 2, the determination criterion data has a first threshold value, a second threshold value, and a third threshold value.

The server 400 acquires the determination reference data 140a from the voice processing device 300, and calculates the initial values of the first threshold value, the second threshold value, and the third threshold value based on the acquired determination reference data. The server 400 transmits the calculated initial value data of the first threshold value, the second threshold value, and the third threshold value to the voice processing device 300. In the following description, the data of the initial values of the first threshold value, the second threshold value, and the third threshold value calculated by the server 400 is referred to as "initial value data".

When the voice processing device 300 receives the initial data from the server 400, the voice processing device 300 updates the determination reference data with the received initial data.

FIG. 10 is a functional block diagram showing the configuration of the voice processing device according to the third embodiment. The voice processing device 300a is connected to the microphone 10 as shown in FIG. The voice processing device 300a includes an AD conversion unit 110, a pitch extraction unit 120a, a power extraction unit 120b, a stress detection unit 130, a storage unit 140, an information presentation unit 160, an estimation unit 210, and an update unit 220. Further, the voice processing device 300a has an upload unit 310 and a download unit 320. Here, the configuration of the voice processing device 300a will be described as an example, but the configurations of the voice processing devices 300b and 300c are the same as the configurations of the voice processing device 300a.

In FIG. 10, the description of the microphone 10, the AD conversion unit 110, the pitch extraction unit 120a, the power extraction unit 120b, the stress detection unit 130, the storage unit 140, and the information presentation unit 160 is the same as that described in the first embodiment. Therefore, the description thereof will be omitted. Since the description of the estimation unit 210 and the update unit 220 is the same as that described in the second embodiment, the description thereof will be omitted.

The upload unit 310 is a processing unit that transmits (uploads) the determination reference data 140a updated by the update unit 220 to the server 400. For example, the upload unit 310 counts the number of calls N between the voice processing device 300a and another voice processing device, and when the number of calls N exceeds the fourth threshold value, the determination reference data 140a is transmitted to the server 400. Send to.

The download unit 320 is a processing unit that receives (downloads) initial value data from the server 400. The download unit 320 updates the determination reference data 140a with the received initial value data. The estimation unit 210 performs processing using the determination reference data 140a updated by the initial value data as the initial value.

It is assumed that the upload unit 310 and the download unit 320 execute data communication with the server 400 via the network 50 by using a communication device (not shown).

FIG. 11A is a functional block diagram showing a server configuration according to the third embodiment. As shown in FIG. 11A, the server 400 has a communication unit 410, a storage unit 420, and a control unit 430.

The communication unit 410 is a processing unit that executes data communication with the voice processing device 300 via the network 50. The control unit 430, which will be described later, exchanges data with the voice processing device 300 via the communication unit 410. The communication unit 410 corresponds to the communication device.

The storage unit 420 has a threshold table 420a. The storage unit 420 corresponds to semiconductor memory elements such as RAM, ROM, and flash memory, and storage devices such as HDD.

The threshold value table 420a is a table that holds the determination reference data 140a transmitted from the voice processing device 300. FIG. 11B is a diagram showing an example of the data structure of the threshold table according to the third embodiment. As shown in FIG. 11B, the threshold table 420a associates the identification information with the determination reference data. The identification information is information that uniquely identifies the voice processing device 300. The judgment standard data is the judgment standard data received from the voice processing device. As described in Examples 1 and 2, the determination reference data 140a includes a first threshold value, a second threshold value, and a third threshold value.

The control unit 430 includes a reception unit 430a, a calculation unit 430b, and a distribution unit 430c. The control unit 430 can be realized by a CPU, an MPU, or the like. The control unit 430 can also be realized by hard-wired logic such as ASIC or FPGA.

The receiving unit 430a is a processing unit that receives the determination reference data 140a from the voice processing device 300. For example, it is assumed that the determination reference data 140a is provided with identification information that identifies the voice processing device 300 that is the source of the determination reference data 140a. The reception unit 430a registers the determination reference data 140a in the threshold table 420a in association with the identification information.

The calculation unit 430b is a processing unit that calculates initial value data based on the threshold table 420a. The calculation unit 430b outputs the calculated initial value data to the distribution unit 430c. An example of the processing of the calculation unit 430b will be described below.

The calculation unit 430b refers to the threshold table 420a and starts the calculation process when the number of records registered in the threshold table 420a is equal to or greater than the fifth threshold. For example, the fifth threshold is set to "3". Since the threshold table 420a shown in FIG. 11B has the determination reference data 140a received from the voice processing devices 300a to 300c (the number of records is 3 or more), the calculation unit 430b executes the calculation process.

An example of the calculation process executed by the calculation unit 430b will be described. The calculation unit 430b calculates the initial value μ1 of the first threshold value by calculating the average value of the first threshold values of each determination reference data 140a. The calculation unit 430b calculates the initial value μ2 of the second threshold value by calculating the average value of the second threshold value of each determination reference data 140a. The calculation unit 430b calculates the initial value μ3 of the third threshold value by calculating the average value of the third threshold value of each determination reference data 140a.

The calculation unit 430b outputs the above initial values μ1 to μ3 as initial value data to the distribution unit 430c.

The distribution unit 430c is a processing unit that transmits the acquired initial value data to the voice processing device 300 when the initial value data is acquired from the calculation unit 430b.

Next, the processing procedure of the voice processing device 300 according to the third embodiment will be described. 12 and 13 are flowcharts showing a processing procedure of the voice processing apparatus according to the third embodiment. As shown in FIG. 12, the AD conversion unit 110 of the voice processing device 300 starts accepting the input voice (step S301). The AD conversion unit 110 performs AD conversion (step S302). The pitch extraction unit 120a of the voice processing device 300 extracts the pitch, and the power extraction unit 120b of the voice processing device 300 extracts the power (step S303).

The pitch extraction unit 120a detects a sound section (step S304). The stress detection unit 130 of the voice processing device 300 accumulates pitch power (step S305). When the pitch power corresponding to the specified number of frames is accumulated (step S306, Yes), the stress detection unit 130 proceeds to step S307. On the other hand, the stress detection unit 130 proceeds to step S301 when the pitch power corresponding to the specified number of frames is not accumulated (steps S306, No).

The stress detection unit 130 calculates the stress value (step S307). The estimation unit 210 of the voice processing device 300 calculates the average power of the sounded section (step S308). When the average power is equal to or higher than the first threshold value (step S309, Yes), the estimation unit 210 shifts to step S310. On the other hand, when the average power is less than the first threshold value (step S309, No), the estimation unit 210 shifts to step S312.

The estimation unit 210 determines whether or not the stress value is equal to or higher than the third threshold value (step S310). When the stress value is equal to or higher than the third threshold value (step S310, Yes), the estimation unit 210 shifts to step S315. When the stress value is less than the third threshold value (step S310, No), the estimation unit 210 causes the information presentation unit 160 to display the first message (step S311), and proceeds to step S315. For example, the first message is "Please move the microphone away from your mouth."

The process proceeds to the description of step S312. The estimation unit 210 determines whether or not the average power is less than the second threshold value (step S312). When the average power is not less than the second threshold value (step S312, No), the estimation unit 210 shifts to step S315. On the other hand, when the average power is less than the second threshold value (step S312, Yes), the estimation unit 210 shifts to step S313.

The estimation unit 210 determines whether or not the stress value is equal to or higher than the third threshold value (step S313). If the stress value is not equal to or higher than the third threshold value (steps S313, No), the estimation unit 210 proceeds to step S315. On the other hand, when the stress value is equal to or higher than the third threshold value (step S313, Yes), the estimation unit 210 causes the information presentation unit 160 to display the second message (step S314), and proceeds to step S315. For example, the second message is "Please bring the microphone a little closer to your mouth."

The update unit 220 of the voice processing device 300 executes the update process (step S315). The update process shown in step S315 corresponds to the process shown in FIGS. 5, 6 and 7 of the second embodiment. The estimation unit 210 stores the stress value and the average power in the storage unit 140 (step S316), and proceeds to step S317 of FIG.

FIG. 13 will be described. The voice processing device 300 determines whether or not the call has ended (step S317). If the call is not completed (step S317, No), the voice processing unit 300 proceeds to step S301 in FIG.

When the call is completed (step S317, Yes), the upload unit 310 of the voice processing device 300 adds 1 to the number of calls N (step S318). When the number of calls N is not equal to or greater than the fourth threshold value (step S319, No), the upload unit 310 ends the process.

On the other hand, when the number of calls N is equal to or greater than the fourth threshold value (step S319, Yes), the upload unit 310 transmits the determination reference data 140a to the server 400 (step S320).

Next, the processing procedure of the server 400 according to the third embodiment will be described. FIG. 14 is a flowchart showing a processing procedure of the server according to the third embodiment. As shown in FIG. 14, the receiving unit 430a of the server 400 receives the determination reference data 140a (step S401). The receiving unit 430a registers the determination reference data 140a in the threshold table 420a (step S402).

The calculation unit 430b of the server 400 determines whether or not the number of records of the determination reference data is equal to or greater than the fifth threshold value based on the threshold value table 420a (step S403). If the number of records of the determination reference data is not equal to or greater than the fifth threshold value (steps S403, No), the calculation unit 430b ends the process.

On the other hand, when the number of records of the determination reference data is equal to or greater than the fifth threshold value (step S403, Yes), the calculation unit 430b calculates the average value for each of the first threshold value to the third threshold value, and the initial values μ1 to μ1 to Specify μ3 (step S404).

The calculation unit 430b generates initial value data (step S405). The distribution unit 430c of the server 400 transmits the initial value data to the voice processing device 300 (step S406). Here, the voice processing device 300 that transmits the initial value data may be a newly introduced voice processing device that has never been used.

Next, the effect of the system according to the third embodiment will be described. The voice processing device 300 notifies the server 400 of the determination reference data 140a after repeatedly executing the update process, and the server 400 generates initial value data based on each determination reference data 140a, and the voice processing device 300 generates initial value data. Notify to. By using the initial value data, the voice processing device 300 can use the more correct first threshold value, second threshold value, and third threshold value as the initial value determination reference data 140a.

In addition, in this Example 3, the following processing is also possible. For example, the server 400 generates initial value data based on the determination reference data 140a of the voice processing devices 300b, 300c, and other voice processing devices, and generates the initial value data when the voice processing device 300a is started. The initial value data is transmitted to the voice processing device 300a. By using the initial value data received from the server 400 from the time of startup, the voice processing device 300a can notify the user of a more correct message without repeatedly executing the above-mentioned update process. In addition, by downloading the initial value to a newly introduced voice processing device that has never been used, the judgment criteria data that has already been updated from the first use can be used, so a more correct message from the first time. Can be notified to the user.

FIG. 15 is a diagram showing an example of the system according to the fourth embodiment. As shown in FIG. 15, this system has voice processing devices 500a to 500l and a server 600. The voice processing devices 500a to 500l and the server 600 are connected to each other via the network 50. Here, as an example, the voice processing devices 500a to 500l are shown, but other voice processing devices may be included. In the following description, the voice processing devices 500a to 500l are collectively referred to as the voice processing device 500 as appropriate.

As an example in the fourth embodiment, the voice processing devices 500a to 500c are arranged in the room 10A. Therefore, the voice processing devices 500a to 500c have similar usage environments. The voice processing devices 500d to 500f are arranged in the room 10B. Therefore, the voice processing devices 500d to 500f have similar usage environments. The voice processing devices 500g to 500i are arranged in the room 10C. Therefore, the voice processing devices 500g to 500i have similar usage environments. The voice processing devices 500j to 500l are arranged in the room 10D. Therefore, the voice processing devices 500j to 500l have similar usage environments.

The voice processing device 500 repeatedly executes the update process in the same manner as the voice processing device 200 described in the second embodiment, and transmits the updated determination reference data 140a to the server 600. As described in Examples 1 to 3, the determination reference data 140a has a first threshold value, a second threshold value, and a third threshold value.

The server 600 acquires the determination reference data 140a from the voice processing device 500, and based on each first threshold value of the acquired determination reference data, is used by a user who has a loud voice as compared with other voice processing devices 500. The voice processing device 500 to be used is specified. The server 600 sends a third message "Please make your voice a little quieter" to the specified voice processing device 500. The voice processing device 500 that has received the third message presents the third message to the user.

The server 600 acquires the determination reference data 140a from the voice processing device 500, and based on each second threshold value of the acquired determination reference data, the server 600 is used by a user having a lower voice than the other voice processing device 500. The voice processing device 500 to be used is specified. The server 600 sends a fourth message "Please make your voice a little louder" to the specified voice processing device 500. The voice processing device 500 that has received the fourth message presents the fourth message to the user.

FIG. 16 is a functional block diagram showing the configuration of the voice processing device according to the fourth embodiment. As shown in FIG. 16, the voice processing device 500a is connected to the microphone 10 as shown in FIG. The voice processing device 500a includes an AD conversion unit 110, a pitch extraction unit 120a, a power extraction unit 120b, a stress detection unit 130, a storage unit 140, an information presentation unit 160, an estimation unit 210, and an update unit 220. Further, the voice processing device 500a has a message receiving unit 510. Here, the configuration of the voice processing device 500a will be described as an example, but the configuration of the voice processing devices 500b to 500l is the same as the configuration of the voice processing device 500a.

In FIG. 16, the description of the microphone 10, the AD conversion unit 110, the pitch extraction unit 120a, the power extraction unit 120b, the stress detection unit 130, the storage unit 140, and the information presentation unit 160 is the same as that described in the first embodiment. Therefore, the description thereof will be omitted. Since the description of the estimation unit 210 and the update unit 220 is the same as that described in the second embodiment, the description thereof will be omitted.

When the message receiving unit 510 receives a message from the server 600 via the communication device, the message receiving unit 510 causes the information presenting unit 160 to present the received message. For example, the message received from the server 600 is the third message or the fourth message as described above.

The update unit 220 counts the number of updates each time the determination reference data 140a is updated. When the number of updates of the determination reference data 140a exceeds a predetermined number of times, the update unit 220 transmits the determination reference data 140a to the server 600 by using the communication device.

FIG. 17 is a functional block diagram showing a server configuration according to the fourth embodiment. As shown in FIG. 17, the server 600 has a communication unit 610, a storage unit 620, and a control unit 630.

The communication unit 610 is a processing unit that executes data communication with the voice processing device 500 via the network 50. The control unit 630, which will be described later, exchanges data with the voice processing device 500 via the communication unit 610. The communication unit 610 corresponds to the communication device.

The storage unit 620 has a threshold table 620a. The storage unit 620 corresponds to semiconductor memory elements such as RAM, ROM, and flash memory, and storage devices such as HDD.

The threshold table 620a is a table that holds the determination reference data 140a transmitted from the voice processing device 500. Since the data structure of the threshold table 620a corresponds to the threshold table 420a described with reference to FIG. 11B, the description thereof will be omitted.

The classification table 620b is a table that holds data of the group to which the voice processing device 500 belongs. FIG. 18 is a diagram showing an example of the data structure of the classification table according to the fourth embodiment. As shown in FIG. 18, the classification table 620b associates the group identification information with the identification information. The group identification information is information that uniquely identifies a group. The identification information is information that uniquely identifies the voice processing device 500.

The voice processing devices 500 classified into the same group have similar usage environments. For example, the voice processing devices 500a to 500c are classified into the same group. The voice processing devices 500d to 500f are classified into the same group. The voice processing devices 500g to 500i are classified into the same group. The voice processing devices 500j to 500l are classified into the same group.

The control unit 630 includes a reception unit 630a, a statistic calculation unit 630b, an outlier extraction unit 630c, and a message transmission unit 630d. The control unit 630 can be realized by a CPU, an MPU, or the like. The control unit 630 can also be realized by hard-wired logic such as ASIC or FPGA.

The receiving unit 630a is a processing unit that receives the determination reference data 140a from the voice processing device 500. For example, it is assumed that the determination reference data 140a is provided with identification information that identifies the voice processing device 500 that is the source of the determination reference data 140a. The receiving unit 630a registers the determination reference data 140a in the threshold table 620a in association with the identification information.

The statistic calculation unit 630b is a processing unit that calculates statistics for the same group based on the threshold table 620a. The statistic calculation unit 630b calculates the average value μ1 of the first threshold value and the standard deviation σ1 of the first threshold value as statistics. Further, the statistic calculation unit 630b calculates the average value μ2 of the second threshold value and the standard deviation σ2 of the second threshold value. The statistic calculation unit 630b outputs the statistic information for each group to the outlier extraction unit 630c.

The statistic calculation unit 630b identifies the identification information of the voice processing device 500 belonging to the same group by referring to the classification table 620b. The statistic calculation unit 630b acquires the determination reference data 140a (first threshold value, second threshold value) of the voice processing device 500 belonging to the same group by comparing the specified identification information with the threshold value table 620a. The statistical value calculation unit 630b calculates the above-mentioned statistic by using the first threshold value and the second threshold value of the voice processing device 500 belonging to the same group.

FIG. 19 is a diagram showing an example of a statistical data structure. As shown in FIG. 19, this statistic associates the group identification information with the first mean, the first standard deviation, the second mean, and the second standard deviation. The group identification information is information that uniquely identifies a group. The first mean value indicates the mean value of each first threshold value of the same group. The first standard deviation indicates the standard deviation of each first threshold in the same group. The second mean value indicates the mean value of each second threshold value of the same group. The second standard deviation indicates the standard deviation of each second threshold in the same group.

The outlier extraction unit 630c uses the statistic, the threshold value table 620a, and the classification table 620b to transmit the first threshold value or the second threshold value (determination reference data 140a) of the outliers to the voice processing device 500. It is a processing unit that extracts each group.

The outlier extraction unit 630c compares the first threshold value of the voice processing device 500 included in the same group with the “first mean value μ1 + 3 × first standard deviation σ1” of the corresponding group. The outlier extraction unit 630c extracts the voice processing device 500 whose first threshold value exceeds “first average value μ1 + 3 × first standard deviation σ1” as the “first outlier”. In the present embodiment, a value 3 times the standard deviation from the average value is set as the threshold value, but the threshold value is not limited to 3 times, and may be set to 2 times or 1 time.

The outlier extraction unit 630c compares the second threshold value of the voice processing device 500 included in the same group with the “second mean value μ2-3 × second standard deviation σ2” of the corresponding group. The outlier extraction unit 630c extracts the voice processing device 500 whose second threshold value is less than the “second mean value μ2-3 × second standard deviation σ2” as the “second outlier”. The outlier extraction unit 630c outputs the identification information of the first outlier and the identification information of the second outlier to the message transmission unit 630d. In the present embodiment, a value 3 times the standard deviation from the average value is set as the threshold value, but the threshold value is not limited to 3 times, and may be set to 2 times or 1 time.

The outlier extraction unit 630c repeatedly executes the above processing for each group, and outputs the identification information of the first outlier and the identification information of the second outlier for each group to the message transmission unit 630d.

It can be said that a user who is talking using the voice processing device 500 corresponding to the identification information of the first detached device is "loud" as compared with other users who have similar usage environments. It can be said that a user who is making a call using the voice processing device 500 corresponding to the identification information of the second detached device has a "low voice" as compared with other users having a similar usage environment.

The message transmission unit 630d is a processing unit that transmits a message to the voice processing device 500 based on the information acquired from the outlier extraction unit 630c. For example, the message transmission unit 630d transmits the third message "Please make your voice a little quieter" to the voice processing device 500 corresponding to the identification information of the first detached device. For example, the message transmission unit 630d transmits the fourth message "Please make your voice a little louder" to the voice processing device 500 corresponding to the identification information of the second detached device.

Next, the processing procedure of the voice processing device 500 according to the fourth embodiment will be described. FIG. 20 is a flowchart showing a processing procedure of the voice processing device according to the fourth embodiment. As shown in FIG. 20, the AD conversion unit 110 of the voice processing device 500 starts accepting the input voice (step S501). The AD conversion unit 110 performs AD conversion (step S502). The pitch extraction unit 120a of the voice processing device 500 extracts the pitch, and the power extraction unit 120b of the voice processing device 500 extracts the power (step S503).

The pitch extraction unit 120a detects a sounded section (step S504). The stress detection unit 130 of the voice processing device 500 accumulates pitch power (step S505). When the pitch power corresponding to the specified number of frames is accumulated (steps S506 and Yes), the stress detection unit 130 shifts to step S507. On the other hand, when the pitch power corresponding to the specified number of frames is not accumulated (step S506, No), the stress detection unit 130 shifts to step S501.

The stress detection unit 130 calculates the stress value (step S507). The estimation unit 210 of the voice processing device 500 calculates the average power of the sounded section (step S508). When the average power is equal to or higher than the first threshold value (step S509, Yes), the estimation unit 210 shifts to step S510. On the other hand, when the average power is less than the first threshold value (step S509, No), the estimation unit 210 shifts to step S512.

The estimation unit 210 determines whether or not the stress value is equal to or higher than the third threshold value (step S510). When the stress value is equal to or higher than the third threshold value (step S510, Yes), the estimation unit 210 shifts to step S515. When the stress value is less than the third threshold value (step S510, No), the estimation unit 210 causes the information presentation unit 160 to display the first message (step S511), and proceeds to step S515. For example, the first message is "Please move the microphone away from your mouth."

The process proceeds to the description of step S512. The estimation unit 210 determines whether or not the average power is less than the second threshold value (step S512). When the average power is not less than the second threshold value (step S512, No), the estimation unit 210 shifts to step S515. On the other hand, when the average power is less than the second threshold value (step S512, Yes), the estimation unit 210 shifts to step S513.

The estimation unit 210 determines whether or not the stress value is equal to or higher than the third threshold value (step S513). When the stress value is not equal to or higher than the third threshold value (step S513, No), the estimation unit 210 proceeds to step S515. On the other hand, when the stress value is equal to or higher than the third threshold value (step S513, Yes), the estimation unit 210 causes the information presentation unit 160 to display the second message (step S514), and proceeds to step S515. For example, the second message is "Please bring the microphone a little closer to your mouth."

The update unit 220 of the voice processing device 500 executes the update process (step S515). The update process shown in step S515 corresponds to the process shown in FIGS. 5, 6 and 7 of the second embodiment. The estimation unit 160 stores the stress value and the average power in the storage unit 140 (step S516).

When the number of updates exceeds a predetermined number, the update unit 220 transmits the determination reference data 140a to the server 600 (step S517). The update unit 220 adds 1 to the number of updates (step S518), and proceeds to step S501.

Next, the processing procedure of the server 600 according to the fourth embodiment will be described. FIG. 21 is a flowchart showing a processing procedure of the server according to the fourth embodiment. As shown in FIG. 21, the receiving unit 630a of the server 600 receives the determination reference data 140a (step S601). The receiving unit 630a registers the determination reference data 140a in the threshold table 620a (step S602).

The statistic calculation unit 630b of the server 600 determines whether or not the number of records of the determination reference data is equal to or greater than the fifth threshold value based on the threshold value data 620a (step S603). When the number of records of the determination reference data is not equal to or greater than the fifth threshold value (step S603, No), the statistic calculation unit 630b ends the process.

On the other hand, when the number of records of the determination reference data is equal to or greater than the fifth threshold value (step S603, Yes), the statistic calculation unit 630b calculates an average value μ for each of the first threshold value and the second threshold value (step). S604). The statistic calculation unit 630b calculates the standard deviation σ for each of the first threshold value and the second threshold value (step S605).

The outlier extraction unit 630c of the server 600 proceeds to step S608 when the first threshold value does not exceed the first mean value μ1 + 3 × first standard deviation σ1 in the judgment reference data 140a (step S606, No). To do.

The outlier extraction unit 630c proceeds to step S607 when the first threshold value exceeds the first mean value μ1 + 3 × first standard deviation σ1 in the determination reference data 140a (step S606, Yes). The message transmission unit 630d of the server 600 transmits a third message "Please make your voice a little quieter" to the corresponding voice processing device 500 (step S607).

The outlier extraction unit 630c ends the process when there is no second threshold value less than the second mean value μ2-3 × second standard deviation σ2 in the determination reference data 140a (step S608, No).

The outlier extraction unit 630c shifts to step S609 when the second threshold value is less than the second mean value μ2-3 × second standard deviation σ2 in the judgment reference data 140a (step S608, Yes). .. The message transmission unit 630d of the server 600 transmits the fourth message “Please make your voice a little louder” to the corresponding voice processing device 500 (step S609).

Next, the effect of the system according to the fourth embodiment will be described. The voice processing device 500 notifies the server 600 of the determination reference data 140a after repeatedly executing the update process. The server 600 calculates a statistic based on each determination reference data 140a, and sends a message to the voice processing device 500 that transmits the first threshold value of the outliers and the voice processing device 500 that transmits the second threshold value of the outliers. Send. For example, since the server 600 transmits a third message to the voice processing device 500 that has transmitted the first threshold value of the outlier, the voice of a user who speaks relatively loudly among the voice processing devices having similar usage environments. It is possible to issue attention to the processing device. Since the server 600 transmits the fourth message to the voice processing device 500 that has transmitted the second threshold value of the outlier, the voice processing device of the user who speaks relatively quietly among the voice processing devices having similar usage environments. Attention can be issued to 500.

In the system according to the fourth embodiment, the voice processing device 500 is divided into groups having similar usage environments, and the third message and the fourth message are transmitted for each group, but the system is limited to this. is not it. In the system according to the fourth embodiment, the voice processing devices 500a to 500l may be grouped into one group and the same processing may be executed.

FIG. 22 is a functional block diagram showing a configuration of the voice processing device according to the fifth embodiment. As shown in FIG. 22, the voice processing device 700 is connected to the microphone 10. The voice processing device 700 includes an AD conversion unit 110, a pitch extraction unit 120a, a power extraction unit 120b, a stress detection unit 130, a storage unit 140, an estimation unit 710, and a gain adjustment unit 720. Of these, the description of the microphone 10, the AD conversion unit 110, the pitch extraction unit 120a, the power extraction unit 120b, the stress detection unit 130, and the storage unit 140 is the same as that described in the first embodiment, and thus the description thereof will be omitted. ..

The estimation unit 710 is a processing unit that estimates the utterance state of the input voice based on the stress value of the input voice, the average power, and the determination reference data 140a. The estimation unit 710 outputs the estimation result to the gain adjustment unit 720.

For example, the estimation unit 710 outputs the first estimation result to the gain adjustment unit 720 when the stress is small and the average power is equal to or higher than the first threshold value. The estimation unit 710 outputs the second estimation result to the gain adjustment unit 720 when “the stress is large” and “the average power is less than the second threshold value”. The process in which the estimation unit 710 compares the high, low, and average power of stress with the first threshold value and the second threshold value is the same as the process of the estimation unit 150 shown in Example 1.

The gain adjusting unit 720 is a processing unit that adjusts the gain of the microphone 10 based on the estimation result of the estimation unit 710. When the gain adjusting unit 720 receives the first estimation result from the estimation unit 710, the gain adjusting unit 720 lowers the gain of the microphone 10. For example, the gain adjusting unit 720 lowers the recording level of the microphone 10 by 3 dB. The first estimation result shows that the current stress is "small" and "the average power is equal to or higher than the first threshold value", and when the stress changes to "large" in the future, the power of the input voice is appropriate. There is a risk of exceeding the power.

When the gain adjusting unit 720 receives the second estimation result from the estimation unit 710, the gain adjusting unit 720 raises the gain of the microphone 10. For example, the gain adjusting unit 720 raises the recording level of the microphone 10 by 3 dB. The second estimation result indicates that the current stress is "large" and "the average power is less than the second threshold", and when the stress changes to "small" in the future, the power of the input voice is appropriate. There is a risk of falling below the power.

Next, the processing procedure of the voice processing device 700 according to the fifth embodiment will be described. FIG. 23 is a flowchart showing a processing procedure of the voice processing device according to the fifth embodiment. As shown in FIG. 23, the AD conversion unit 110 of the voice processing device 700 starts accepting the input voice (step S701). The AD conversion unit 110 performs AD conversion (step S702). The pitch extraction unit 120a of the voice processing device 700 extracts the pitch, and the power extraction unit 120b of the voice processing device 700 extracts the power (step S703).

The pitch extraction unit 120a detects a sounded section (step S704). The stress detection unit 130 of the voice processing device 700 accumulates pitch power (step S705). When the pitch power corresponding to the specified number of frames is accumulated (step S706, Yes), the stress detection unit 130 shifts to step S707. On the other hand, when the pitch power corresponding to the specified number of frames is not accumulated (step S706, No), the stress detection unit 130 shifts to step S701.

The stress detection unit 130 calculates the stress value (step S707). The estimation unit 710 of the voice processing device 700 calculates the average power of the sounded section (step S708). When the average power is equal to or higher than the first threshold value (step S709, Yes), the estimation unit 710 shifts to step S710. On the other hand, when the average power is less than the first threshold value (step S709, No), the estimation unit 710 shifts to step S712.

The estimation unit 710 determines whether or not the stress value is equal to or higher than the third threshold value (step S710). When the stress value is equal to or higher than the third threshold value (step S710, Yes), the estimation unit 710 shifts to step S701. When the stress value is less than the third threshold value (step S710, No), the estimation unit 710 shifts to step S711. The gain adjusting unit 720 of the voice processing device 700 lowers the voice level of the microphone 10 by 3 dB (step S711), and shifts to step S701.

The process proceeds to the description of step S712. The estimation unit 710 determines whether or not the average power is less than the second threshold value (step S712). When the average power is not less than the second threshold value (step S712, No), the estimation unit 710 shifts to step S701. On the other hand, when the average power is less than the second threshold value (step S712, Yes), the estimation unit 710 shifts to step S713.

The estimation unit 710 determines whether or not the stress value is equal to or higher than the third threshold value (step S713). When the stress value is not equal to or higher than the third threshold value (step S713, No), the estimation unit 710 proceeds to step S701. On the other hand, when the stress value is equal to or higher than the third threshold value (step S713, Yes), the estimation unit 710 shifts to step S714. The gain adjusting unit 720 raises the sound level of the microphone 10 by 3 dB (step S714), and shifts to step S701.

Next, the effect of the voice processing device 700 according to the fifth embodiment will be described. The voice processing device 700 estimates the utterance state and adjusts the gain of the microphone 10 based on the transition result. As a result, the volume of the input voice of the user can be kept at an appropriate volume in consideration of the psychological situation of the user, and the call of each user can be kept comfortable.

Next, an example of a computer hardware configuration that realizes the same functions as the voice processing devices 100, 200, 300, 500, and 700 shown in the above embodiment will be described. FIG. 24 is a diagram showing an example of a computer hardware configuration that realizes a function similar to that of a voice processing device.

As shown in FIG. 24, the computer 800 includes a CPU 801 that executes various arithmetic processes, an input device 802 that receives data input from a user, and a display 803. Further, the computer 800 has a reading device 804 that reads a program or the like from a storage medium, and an interface device 805 that exchanges data with another computer via a wired or wireless network. For example, the interface device 805 is connected to a communication device or the like. The computer 800 is connected to the microphone 806. Further, the computer 800 has a RAM 807 that temporarily stores various information and a hard disk device 808. Then, each of the devices 801-808 is connected to the bus 809.

The hard disk device 808 has an extraction program 808a, a stress detection program 808b, an estimation program 808c, and an update program 808d. Further, the hard disk device 808 has an upload / download program 808e, a reception program 808f, a presentation program 808g, and a gain adjustment program 808h. The CPU 801 reads out the extraction program 808a, the stress detection program 808b, the estimation program 808c, and the update program 808d and deploys them in the RAM 807. The CPU 801 reads the upload / download program 808e, the reception program 808f, the presentation program 808g, and the gain adjustment program 808h and deploys them in the RAM 807.

The extraction program 808a functions as an extraction process 807a. The stress detection program 808b functions as the stress detection process 807b. The estimation program 808c functions as an estimation process 807c. Update 808d functions as update process 807d. The upload / download program 808e functions as an upload / download process 807e. The receiving program 808f functions as a receiving process 807f. The presentation program 808g functions as a presentation process 807g. The gain adjustment program 808h functions as a gain adjustment process 807h.

The processing of the extraction process 807a corresponds to the processing of the pitch extraction unit 120a and the power extraction unit 120b. The process of the stress detection process 807b corresponds to the process of the stress detection unit 130. The processing of the estimation process 807c corresponds to the processing of the estimation units 150, 210, and 710. The process of the update process 807d corresponds to the process of the update unit 220. The processing of the upload / download process 807e corresponds to the processing of the upload unit 310 and the download unit 320. The processing of the receiving process 807f corresponds to the processing of the message receiving unit 510. The processing of the presentation process 807g corresponds to the processing of the information presentation unit 160. The processing of the gain adjusting process 807h corresponds to the processing of the gain adjusting unit 720.

The programs 808a to 808h do not necessarily have to be stored in the hard disk device 808 from the beginning. For example, each program is stored in a "portable physical medium" such as a flexible disk (FD), a CD-ROM, a DVD disk, a magneto-optical disk, or an IC card inserted into the computer 800. Then, the computer 800 may read and execute each program 808a to 808h.

The following additional notes will be further disclosed with respect to the embodiments including each of the above embodiments.

(Appendix 1) To the computer
Extract the pitch frequency and frequency power from the input voice,
A determination result is output as to whether or not the condition that the value based on the pitch frequency and the frequency power is equal to or higher than a predetermined threshold value is satisfied.
A voice processing program characterized in that a process of estimating the utterance state of the input voice is executed based on the relationship between the determination result and the average power of the frequency power.

(Appendix 2) The estimation result of the vocalization state is notified, it is determined whether or not the vocalization state is improved, and if it is improved, the judgment standard or the threshold value used by the process of estimating the vocalization state is correctly answered. If the data is not improved or if there is no change, the incorrect answer data is reduced by using the judgment criterion or the threshold value as incorrect answer data and using the correct answer data and the incorrect answer data as training data. The voice processing program according to Appendix 1, wherein the computer further executes a process of updating the determination criterion or the threshold value.

(Appendix 3) The determination criteria are collected from a server that collects the learning data from a plurality of terminals connected to the network and generates an initial value of the determination criterion and an initial value of the threshold value based on the collected learning data. The initial value of the above threshold value and the initial value of the threshold value are received, the initial value of the determination criterion is set to the initial value of the determination criterion used in the process of estimating the vocalization state, and the initial value of the threshold value is the determination process. The voice processing program according to Appendix 2, wherein the computer further executes the process of setting the initial value of the threshold value.

(Appendix 4) The server collects a plurality of threshold values updated based on the learning data from the plurality of terminals, and compares the collected plurality of threshold values with each other to identify a user with a loud voice. , Notify the terminal used by the specified user of the alarm,
The voice processing program according to Appendix 3, wherein when the alarm is received, the computer further executes a process of outputting the alarm.

(Appendix 5) The server classifies the plurality of terminals into groups according to the usage environment, and compares a plurality of thresholds acquired from a plurality of terminals classified into groups having similar usage environments to make a voice voice. The voice processing program according to Appendix 4, which is characterized by identifying a large user.

(Appendix 6) When the value based on the pitch frequency and the frequency power is less than the third threshold value and the average power of the frequency power is equal to or more than the first threshold value, or the value based on the pitch frequency and the frequency power is the first. When the average power of the frequency power is less than or equal to the second threshold value of 3 threshold values or more, any one of Supplementary notes 1 to 5 is characterized in that the computer is further executed to add a correction gain to the input voice. The voice processing program described in one.

(Appendix 7) A voice processing method executed by a computer.
Extract the pitch frequency and frequency power from the input voice,
The determination result is determined whether or not the condition that the value based on the pitch frequency and the frequency power is equal to or higher than a predetermined threshold value is satisfied.
A voice processing method characterized by executing a process of estimating a vocalization state of the input voice based on the relationship between the determination result and the average power of the frequency power.

(Appendix 8) The estimation result of the vocalization state is notified, it is determined whether or not the vocalization state is improved, and if it is improved, the determination criterion or the threshold value used in the process of estimating the vocalization state is correctly answered. If the data is not improved or if there is no change, the incorrect answer data is reduced by using the judgment criterion or the threshold value as incorrect answer data and using the correct answer data and the incorrect answer data as learning data. The voice processing method according to Appendix 7, wherein the process of updating the determination criterion or the threshold value is further executed.

(Appendix 9) The determination criterion is collected from a server that collects the learning data from a plurality of terminals connected to the network and generates an initial value of the determination criterion and an initial value of the threshold value based on the collected learning data. The initial value of the above threshold value and the initial value of the threshold value are received, the initial value of the determination criterion is set to the initial value of the determination criterion used in the process of estimating the vocalization state, and the initial value of the threshold value is the determination process. 8. The voice processing method according to Appendix 8, wherein the process of setting the initial value of the threshold value is further executed.

(Appendix 10) The server collects a plurality of threshold values updated based on the learning data from the plurality of terminals, and compares the collected plurality of threshold values with each other to identify a user with a loud voice. , Notify the terminal used by the specified user of the alarm,
The voice processing method according to Appendix 9, wherein when the alarm is received, a process of outputting the alarm is further executed.

(Appendix 11) The server classifies the plurality of terminals into groups according to the usage environment, and compares a plurality of threshold values acquired from a plurality of terminals classified into groups having similar usage environments to make a voice voice. The voice processing method according to Appendix 10, wherein a large user is specified.

(Appendix 12) When the value based on the pitch frequency and the frequency power is less than the third threshold value and the average power of the frequency power is equal to or more than the first threshold value, or the value based on the pitch frequency and the frequency power is the first. When the average power of the frequency power is less than or equal to the second threshold value of 3 threshold values or more, any one of Supplementary notes 7 to 11 is characterized in that the computer is further executed to add a correction gain to the input voice. The voice processing method described in one.

(Appendix 13) An extraction unit that extracts the pitch frequency and frequency power from the input voice,
The determination result is determined whether or not the condition that the pitch frequency and the value based on the frequency power satisfy the predetermined threshold value or more is satisfied, and the input voice is based on the relationship between the determination result and the average power of the frequency power. The estimation unit that estimates the vocalization state of
A voice processing device characterized by having.

(Appendix 14) The estimation result of the vocalization state is notified, it is determined whether or not the vocalization state is improved, and if it is improved, the determination criterion or the threshold value used in the process of estimating the vocalization state is correctly answered. If the data is not improved or if there is no change, the incorrect answer data is reduced by using the judgment criterion or the threshold value as incorrect answer data and using the correct answer data and the incorrect answer data as learning data. The voice processing apparatus according to Appendix 13, further comprising an update unit for updating the determination criterion or the threshold value.

(Appendix 15) The determination criterion is collected from a server that collects the learning data from a plurality of terminals connected to the network and generates an initial value of the determination criterion and an initial value of the threshold value based on the collected learning data. The initial value of the above threshold value and the initial value of the threshold value are received, the initial value of the determination criterion is set to the initial value of the determination criterion used in the process of estimating the vocalization state, and the initial value of the threshold value is the determination process. The voice processing apparatus according to Appendix 14, further comprising a download unit for setting the initial value of the threshold value of the above.

(Appendix 16) The server collects a plurality of threshold values updated based on the learning data from the plurality of terminals, and compares the collected plurality of threshold values with each other to identify a user with a loud voice. , Notify the terminal used by the specified user of the alarm,
The voice processing device according to Appendix 15, further comprising a message receiving unit that outputs an alarm when the alarm is received.

(Appendix 17) The server classifies the plurality of terminals into groups according to the usage environment, and compares a plurality of threshold values acquired from a plurality of terminals classified into groups having similar usage environments to make a voice voice. The voice processing device according to Appendix 16, wherein a large user is identified.

(Appendix 18) When the value based on the pitch frequency and the frequency power is less than the third threshold value and the average power of the frequency power is equal to or more than the first threshold value, or the value based on the pitch frequency and the frequency power is the first. When the average power of the frequency power is less than or equal to the second threshold value of 3 threshold values or more, any one of Appendix 13 to 17, further comprising a gain adjusting unit for adding a correction gain to the input sound. The voice processing device described in 1.

10 Microphone 100, 200, 300a, 300b, 300c, 500a, 500b, 500c, 500d, 500e, 500f, 500g, 500h, 500i, 500j, 500k, 500l, 700 Audio processing device 110 AD conversion unit 120a Pitch extraction unit 120b Power Extraction unit 130 Stress detection unit 140 Storage unit 140a Judgment standard data 160 Information presentation unit 220 Update unit 310 Upload unit 320 Download unit 400,600 Server 510 Message reception unit 720 Gain adjustment unit

Claims (8)

On the computer
Extracts the frequency power and the pitch frequency from the input speech,
Wherein a frequency power and a value based on the pitch frequency, the value indicating the user's stress having issued the input speech is determined satisfying whether the determination result equal to or greater than a predetermined threshold value,
A voice processing program characterized in that a process of estimating the utterance state of the input voice is executed based on the relationship between the determination result and the average power of the frequency power. The estimation result of the vocalization state is notified, it is determined whether or not the vocalization state is improved, and if it is improved, the judgment standard or the threshold value used in the process of estimating the vocalization state is used as correct answer data, and the improvement is made. If it is not done or there is no change, the judgment criterion or the threshold value is used as incorrect answer data, and the correct answer data and the incorrect answer data are used as training data so that the incorrect answer data is reduced. The voice processing program according to claim 1, wherein the computer further executes a process of updating the determination standard or the threshold value. From a server that collects the training data from a plurality of terminals connected to the network and generates an initial value of the determination criterion and an initial value of the threshold value based on the collected learning data, the initial value of the determination criterion and The initial value of the threshold value is received, the initial value of the determination criterion is set to the initial value of the determination criterion used in the process of estimating the vocalization state, and the initial value of the threshold value is set to the threshold value of the determination process. The voice processing program according to claim 2, wherein the computer further executes the process of setting the initial value. The server collects a plurality of threshold values updated based on the learning data from the plurality of terminals, and compares the collected plurality of threshold values with each other to identify and identify a user with a loud voice. Notifies the terminal used by the person of the alarm and
The voice processing program according to claim 3, wherein when the alarm is received, a computer further executes a process of outputting the alarm. The server classifies the plurality of terminals into groups according to the usage environment, and compares a plurality of thresholds acquired from the plurality of terminals classified into groups having similar usage environments to obtain a loud user. The voice processing program according to claim 4, wherein the voice processing program is specified. Wherein when the frequency power and a value based on the pitch frequency is the average power of the third threshold value less than and the frequency power is equal to or larger than the first threshold, or a value based on the frequency power and the pitch frequency is more than the third threshold value Further, when the average power of the frequency power is less than the second threshold value, any one of claims 1 to 5, wherein the computer further executes a process of adding a correction gain to the input voice. The voice processing program described in. A computer-executed voice processing method
Extracts the frequency power and the pitch frequency from the input speech,
Wherein a frequency power and a value based on the pitch frequency, the value indicating the user's stress having issued the input speech is determined satisfying whether the determination result equal to or greater than a predetermined threshold value,
A voice processing method characterized by executing a process of estimating a vocalization state of the input voice based on the relationship between the determination result and the average power of the frequency power. An extractor that extracts frequency power and pitch frequency from input audio,
Wherein a frequency power and a value based on the pitch frequency, the value indicating the user's stress having issued the input speech is determined satisfying whether the determination result equal to or greater than a predetermined threshold value, the determination An estimation unit that estimates the vocalization state of the input voice based on the relationship between the result and the average power of the frequency power.
A voice processing device characterized by having.

Images