JP2010259691A - Stress measuring device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a stress device for measuring stress during a conversation through mobile communication devices and measuring the degrees of both speakers' stress caused by the conversation. <P>SOLUTION: The stress measuring device includes a voice input section for inputting voices of both speakers in conversation to make voice signals; a voice analyzing section for discriminating vocal parts out of the voice signals from the voice input section and computing stress indexes from the frequencies and amplitude of the discriminated vocal parts; and a voice stress determining section for determining the stress values of the speakers based on the stress indexes. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a technique for measuring stresses of both parties using voice signals of a talking party. More specifically, the present invention relates to a technique for extracting pitch and amplitude from voice data and measuring stresses of both parties based on the voice data.

  In recent years, as society becomes more complex, the mental stress experienced by humans has increased. This mental stress is a factor of physical symptoms such as psychosomatic disorders and mental illnesses such as depression and interferes with daily life. Various techniques are known for measuring this mental stress.

  Many of the stress determination methods use human biological information for measurement. Commonly used biological information includes an increase in body temperature, a change in skin electrical resistance, a change in blood pressure, a change in blood flow, and the like.

  As an example, FIG. 8 shows a portable communication device that performs stress measurement based on skin electrical resistance. This portable communication device includes an electrode 53 for measuring skin resistance on a portable terminal. The test subject puts on the stress measurement power supply 54, makes two fingers contact the electrode 54, and estimates the stress by measuring the potential. The result is displayed on the liquid crystal display 51, and the relaxation information according to the stress determination result is provided. Relax information is transmitted from the storage means provided in the center server connected by wireless communication. As relaxation information, advice text, music, video, exercise video, etc. for relaxing are provided (for example, refer to Patent Document 1).

JP 2003-153905 A

  However, although the conventional configuration discloses a technique for relieving the stress of the subject by using a mobile communication device, the stress of both speakers speaking using the mobile terminal in real time is disclosed. It cannot be measured. We often experience the stress of conversation every day, especially in conversation over the phone, the pleasant discomfort due to the conversation is more intensely stressed by the speaker because each other's facial expressions are not visible. That is, the conventional configuration has a problem that it is impossible to measure both stresses during conversation by the portable communication device.

  An object of the present invention is to solve the above-described conventional problems, and to provide a stress measuring device capable of measuring a stress generated by a speaker having a conversation using a portable communication device. .

  In order to solve the above-described conventional problems, a stress measuring device according to the present invention includes a voice input unit that inputs voices of both speakers during a conversation to generate a voice signal, and a frequency of the voice signal from the voice input unit. And a voice analysis unit that calculates a stress index from the amplitude, and a voice stress determination unit that determines a stress value of a speaker based on the stress index.

  According to the stress measuring apparatus of the present invention, by using both voice data during conversation, it is possible to determine the stress of both talkers based on the voice data.

Schematic diagram about frequency of voice waveform Schematic diagram of audio signals of conversation A and conversation B during conversation Schematic diagram of amplitude of speech waveform The block diagram which shows the function structure of the stress measuring apparatus 1 which concerns on embodiment of this invention. The block diagram which shows the functional structure of the audio | voice A analysis part 4 and the functional structure of the audio | voice B analysis part 11 of the stress measuring device 1 which concerns on embodiment of this invention. The flowchart which shows operation | movement of the stress measuring apparatus 1 which concerns on embodiment of this invention. The flowchart which shows the operation | movement of the frequency analysis of the stress measuring device 1 which concerns on carrying of this invention. Configuration diagram of a portable communication device for conventional stress measurement

  Hereinafter, embodiments of the stress measuring apparatus of the present invention will be described in detail with reference to the drawings.

(Embodiment 1)
The stress measuring apparatus according to the present invention measures the stress of both parties using two stress parameters X _An and Y _An obtained from the frequency and amplitude of the voices of both parties. Details will be described below.

First, a method for calculating the stress parameter X _An obtained from the sound frequency will be described. FIG. 1 is a diagram schematically showing a change in audio frequency due to stress. FIG. 1A is a schematic diagram of a speech waveform during conversation, and the pitch obtained from the speech waveform is shown in FIGS. FIG. 1B shows a speech waveform with a normal pitch (no stress), and the absolute value thereof is almost constant, although there are individual differences. However, the pitch changes when stressed during conversation. That is, the voice pitch when not stressed is a normal pitch as shown in FIG. 1 (b), but when stressed, the pitch is changed as shown in FIGS. 1 (c) and 1 (d). Change. By using this, the stress of the speaker during the conversation can be determined from the voice pitch during the conversation.

In FIG. 2, the schematic diagram of the audio | voice signal when the conversation person A and the conversation person B are talking is shown. When attention is paid to the voice signals of both of the talkers A and B, it is divided into a non-speaking part and a speaking part. Therefore, the utterance part between the non-voicing part and the non-voicing part is set as a voice unit. Taking the conversation person A as an example, the speech units are..., S _An-1 , S _An , S _{An + 1,.} In this embodiment, if there is no voice for 0.5 seconds or longer, the voiceless part is determined. Hereinafter, a method for determining the stress during the conversation from the change in the frequency of the voice unit will be described by taking the conversation person A as an example.

First, an average frequency in the nth speech unit S _An is defined as an average frequency f _An . The calculation of the average frequency f _An will be described. Since the frequency is the number of repetitions per second in the vibration waveform of the audio signal, a value that is half the number of times that the average value of the audio amplitude is crossed per second (average value intersection number) may be obtained. In this embodiment, the frequency per second in the n-th voice unit S _An is extracted, and the average value is set as the average frequency f _An . The amount of stress is determined based on a difference value Δf _An with respect to the frequency in the previous speech unit S _An-1 . Δf _An is expressed by the number (1).

At this time, the frequency difference must be evaluated as an absolute value. This is because the voice level of a human may be low due to the stress state, and conversely, the voice may be high. Considering both of these, it is important to determine the frequency difference Δf _An by an absolute value.

Subsequently, the ratio of the obtained difference value Δf _An to the average frequency of the voice unit S _An is calculated. That is, the ratio of how much the frequency has changed with respect to the average frequency of each voice unit of the conversation person A is obtained. This ratio is expressed by the number (2) as the stress parameter X _An .

Next, a method for calculating the stress parameter Y _An based on amplitude, which is another parameter, will be described. FIG. 3 shows the voice waveform of the speaker when talking. Since the amplitude of this speech waveform is vibration energy, it represents the volume of sound produced by the speaker. That is, the volume is larger as the amplitude is larger, and the volume is smaller as the amplitude is smaller. When the person does not feel stress, the amplitude of the voice is constant as shown in FIG. However, when stress is applied during conversation, the volume gradually increases, and the amplitude gradually increases as shown in FIG. Or, depending on the stress, the voice suddenly increases, so the amplitude increases rapidly as shown in FIG.

Using this property, stress during conversation can be determined from the amplitude of the speech waveform. This will be specifically described below. Similar to the frequency analysis, the voice signal during the conversation of the conversation person A is divided into a non-speech part and a speech part. As shown in FIG. 2, voice units..., S _An-1 , S _An , S _{An + 1,.} The average amplitude in the nth speech unit S _An is defined as an average amplitude R _An . The calculation of the average amplitude R _An will be described. Since the amplitude is the absolute value of the displacement in one cycle of the audio signal, the absolute value of the maximum displacement while crossing the average value of the audio amplitude twice may be obtained. In the present embodiment, the amplitude for each period in the n-th speech unit S _An is extracted, and the average is taken as the average amplitude R _An . The stress of the talker A is determined using the difference from the amplitude in the previous speech unit S _An-1 . The difference in amplitude is represented by ΔR _An (3).

Compared with the average amplitude of the conversation person A, a ratio of how much the amplitude difference has changed is obtained. The ratio is expressed by the number (4) as the stress parameter Y _An .

Since the two stress parameters X _An and Y _An obtained as described above are the rate of change for each voice unit, the average of the sum of these rates of change is used to evaluate the stress of the talker A. Need to ask. In the present invention, this value is defined as a stress index Z _A as shown in Equation (5), and this index is used as the stress value of the conversation person A.

Here, N is the total number of audio units. In this way, the stress index Z _A is obtained, and if Z _A > 0, the conversation person A is in a stressed state, and if Z _A <0, it is determined that the stress is not applied.

In (Equation 5), the two stress parameters X _An and Y _An are handled in the same way, but they may be weighted. The calculation formula of the stress index Z _A when the weighting coefficient of the stress parameter X _An is a and the weighting coefficient of the stress parameter Y _An is b is shown in Equation (6).

Similar to the stress determination of the conversation person A, the stress determination of the conversation person B can be determined based on the frequency and amplitude of the vibration waveform of the voice signal. The stress parameter X _Bn is obtained based on the voice unit of the voice of the conversation person B as S _Bn and the average frequency as f _Bn . Further, the stress parameter Y _Bn is obtained based on the average amplitude R _Bn . Z _B is obtained from the average of these stress parameters, and the stress is determined.

  Next, an example of stress determination will be shown, and a determination method using a stress index will be described. Here, FIG. 3C is taken as an example, and a case is considered in which the conversationer A is stressed by conversation and the voice temporarily becomes louder.

Usually, the audible band of a voice that can be heard by humans is 20 Hz to 20000 Hz, and in Japanese conversation, a frequency band from 150 Hz to 1500 Hz is used centering on 1000 Hz. The case where stress is applied to the talking conversation person A and the voice becomes high is taken as an example. Table 1 shows the relationship between the number n of the voice unit S _An in that case, the average frequency f _n , the frequency difference Δf _n of the previous voice unit, and the stress parameter X _An .

In this example, the number (sound unit number) n indicating the voice unit is between 2 and 3, and the average frequency f _An is changed between 7 and 8. Here, the average frequency f _A2 of the voice unit number 2 is 1000 Hz, but the average frequency f _A3 of the voice unit number 3 is changed to 1500 Hz. Therefore, if the number (2) is used, the stress parameter X _An can be obtained as X _An = (| 1500−1000 |) /1125=0.44. In this embodiment, Δf _An / n of the denominator of the stress parameter X _An is the average of the average frequencies f _An from n = 1 to 4.

Subsequently, a stress parameter Y _An based on the amplitude is obtained. Daily conversation is about 60 dB and screaming is 100 dB. Therefore, the volume range when talking is 60 dB to 100 dB. Table 2 shows the amount of change in amplitude for each voice unit number when the same conversation as in Table 1 is made. The relationship between the number n of the voice unit S _An , the average amplitude R _n , the difference ΔR _An in amplitude from the previous voice unit, and the stress parameter Y _An is shown.

In this example, the sound unit number (sound unit number) n is between 2 and 3, between 3 and 4, between 4 and 5, between 5 and 6, and between 6 and 7. Represents that the average amplitude R _An has changed. Since the average amplitude R _A2 of the voice unit number 2 changes to 60 dB and the average amplitude R _A3 of the voice unit number 3 changes to 100 dB, the stress parameter Y _An is represented by Y _An = (100−60) / 70 from Equation (4). = 0.57. In this embodiment, ΔR _An / n of the denominator of the stress parameter Y _An
Is the average of the average amplitude R _An from n = 1 to 4.

From the stress parameter X _An and the stress parameter Y _An obtained as described above, the stress index Z _A is obtained using the number (5). The sum of the stress parameter X _An and the stress parameter Y _An shown in Table 1 and Table 2 is obtained, and the average is obtained with the total number of voice units N = 10. Then, the stress index Z _A is obtained as 0.17. Since the stress index is Z _A > 0, it can be determined that the conversation person A feels stress.

  Details of the configuration of the stress measurement apparatus using the stress measurement method described above will be described. FIG. 4 is a block diagram showing a functional configuration of the stress measuring apparatus according to the embodiment of the present invention.

First, the voice of the conversation person A is input to the voice A input unit 2 and then temporarily stored in the voice A buffer 3. The audio signal stored in the audio A buffer 3 is transferred to the audio A analysis unit 4 as needed. The voice A analysis unit 4 extracts a voice unit from the voice signal of the conversation person A, calculates a stress parameter X _An based on the frequency and a stress parameter Y _An based on the amplitude, and uses the calculated parameters as the voice of the voice unit. It is stored in the voice A data storage unit 8 according to the number. For this reason, the capacity of the voice A storage unit 8 needs to be large enough to record all data during the conversation. This may be determined by calculating from the conversation time and the amount of data per unit time of the voice signal. When the conversation person determines that the conversation is over, or when it is desired to make a stress determination, a determination command signal is sent to the voice A stress determination unit 17 from an external instruction circuit (not shown). At this time, the voice A stress determination unit 17 requests that all data stored in the voice A data storage unit 8 be transferred to the voice A stress determination unit 17. The voice A stress determination unit 17 determines the stress of the talker _A by calculating a stress index Z _A using the transferred data.

Similarly to the voice A, the voice B of the conversation person B is also a stress index using the voice B input unit 9, the voice B buffer 10, the voice B analysis unit 11, the voice B data storage unit 15, and the voice B stress determination unit 18. Z _B is calculated to determine the stress of Talker B. The display unit 16 displays the data of the stress determination results of the voice A and the voice B together. Display means such as a liquid crystal display may be used for the display unit 16. When incorporating the stress measuring device 1 of the present invention into a mobile phone, a display unit of the mobile phone may be used.

Next, details of the voice A analysis unit 4 will be described with reference to FIG. The voice A analyzer 4 includes a voice A voice unit extractor 5, a voice A frequency analyzer 6, a voice A amplitude analyzer 7, a voice A stress parameter X _An calculator 19, and a voice A stress parameter X _An calculator. Is done.

  First, the sound temporarily stored in the sound A buffer 3 is transferred to the sound A sound unit extraction unit 5, where the sound signal is divided into sound units and a sound number is added. The divided audio units are transferred to the audio A frequency analysis 6 and the audio A amplitude analysis 7.

The voice A frequency analysis unit 6 extracts the frequency of the transferred voice unit. Further, the average frequency f _n is calculated, and a difference Δf _n with the frequency of the voice unit having the previous voice number is obtained. The voice A stress parameter X _An calculation unit calculates the stress parameter X _An according to the equation (2) using the average frequency and the frequency difference Δf _n obtained by the voice A frequency analysis unit.

Further, the voice A amplitude analysis unit 7 calculates an average amplitude R _n in the voice unit and calculates a difference ΔR _n from the average amplitude in the previous voice unit. Using the amplitude data, the voice A stress parameter Y _An calculating unit calculates the stress parameter Y _An .

The calculated stress parameters X _An and Y _An are temporarily recorded by the voice A data storage unit 8.

FIG. 5B is a block diagram illustrating a detailed functional configuration of the voice B analysis unit 11. Similarly to the voice A analysis unit 4, a voice unit is extracted by the voice B voice unit extraction unit 12. The voice B frequency analysis unit 13 analyzes the frequency of the voice unit, and the voice B stress parameter X _B calculation unit 21 calculates the stress parameter Y _Bn . Further, the voice B amplitude analysis unit 13 analyzes the amplitude of the voice unit, and the voice B stress parameter Y _Bn calculation unit 21 calculates the stress parameter Y _Bn .

  The operation of the stress measuring apparatus 1 according to the embodiment of the present invention will be described using FIG. FIG. 6 is a flowchart for explaining the operation of the stress measuring apparatus 1 according to the embodiment of the present invention.

When the voice of the talker A is input to the voice A input unit 2, a voice signal is cut out for each certain section. Then, numbered as phonetic units S _n to audio signal clipping. The audio unit is cut out based on the presence or absence of a non-voiced portion of 0.5 seconds or longer. (Step S1)
Next, the frequency is extracted from the voice unit S _n and the average frequency f _n is calculated. Then, a difference Δf _n with respect to the average frequency f _n−1 in the previous speech unit is obtained. Further, the ratio of how much the frequency has changed is obtained with respect to the average frequency in the entire conversation of the conversation person A. The stress parameter X _An representing the ratio is obtained according to the number (2). (Step S2)
Further, the amplitude is extracted from the voice unit S _n and the average amplitude R _n is calculated. Then, a difference ΔR _n with respect to the average amplitude R _n-1 in the previous speech unit is obtained. Furthermore, the ratio of how much the frequency has changed with respect to the average amplitude in conversation is obtained. The parameter Y _Bn representing the ratio is obtained according to the equation (4). (Step S3)
The stress parameter X _An based on these frequencies and the stress parameter Y _An based on the amplitude are temporarily stored. (Step S4)
Check if the conversation is over. If the conversation has not ended, step S2 (frequency analysis), step S3 (amplitude analysis), and step S4 (frequency data / amplitude data recording) are repeated. (Step S5)
If the conversation is over, a stress determination is performed. The stress determination is performed based on the stress parameter X _An and the stress parameter Y _An . The stress index Z _A is obtained from the equation (5). If Z _A > 0, it is determined that the talker A feels stress, and if Z <0, the stress is not felt. (Step S6)
The result of the stress determination is output to a display unit such as a display. The stress evaluation result may be displayed as a numerical value, or may be displayed as an illustration or a color. (Step S7)
Next, details of the frequency analysis process performed in step S2 will be described. FIG. 7A is a flowchart for explaining the operation of frequency analysis.

The voice signal is divided into a voiced part and a non-voiced part in the voice A voice unit extraction unit 5 in advance, and is set as a voice unit S _An .

Starting from n = 1, n-1 = 0. The frequency in the 0th audio unit S _A0 is extracted. And the average of the frequency is calculated | required and it is set as average frequency _fA0 . (Step S21)
When a voice is input to the voice A input unit, the input voice is converted into voice units as needed. Following the audio unit S _A0 , the audio unit S _A1 is extracted. An average frequency corresponding to the voice unit S _A1 is calculated to obtain an average frequency f _A1 . (Step S22)
Subsequently, the difference between the average frequencies f _A0 and f _A1 calculated in step S21 and step S22 is calculated. This frequency difference Δf _A1 is calculated as an absolute value represented by the number (1). (Step S23)
Further, a stress parameter X _A1 is calculated based on the frequency difference Δf _A1 . The stress parameter X _A1 is obtained using the number (2). (Step S24)
Here, it is confirmed whether or not the conversation has ended. If the conversation has not ended, then n = 1 and steps S21 to S24 are performed. As long as voice is input to the voice A input unit and converted into voice units, step S21 to step S24 are repeated with n = n + 1.

  When the conversation ends and no voice is input to the voice A input unit, step S2 of the frequency analysis ends.

  Next, details of the amplitude analysis process performed in step S3 will be described. FIG. 7B is a flowchart for explaining the operation of amplitude analysis.

Similar to the frequency analysis (step S2), it starts from n = 1 and n-1 = 0. The amplitude in the 0th audio unit S _A0 is extracted. Then, the average amplitude is calculated to determine the average frequency R _A0 . (Step S31)
An average amplitude R _A1 corresponding to the first audio unit S _A1 of the audio unit S _A0 is obtained. (Step S32)
Subsequently, the difference between the average amplitudes R _A0 and R _A1 calculated in step S31 and step S32 is calculated. This amplitude difference ΔR _A1 is calculated using the number (3). (Step S33)
Further, the stress parameter Y _A1 is calculated based on the amplitude difference ΔR _A1 . This stress parameter Y _A1 is obtained using the equation (4). (Step S34)
Here, it is confirmed whether or not the conversation has ended. If the conversation has not ended, then n = 1 and steps S31 to S34 are performed. As long as voice is input to the voice A input unit and converted into voice units, step S31 to step S34 are repeated with n = n + 1. When the conversation ends and no voice is input to the voice A input unit, the amplitude analysis step S3 ends.

  As described above, in the first embodiment of the present invention, it is possible to determine the stress of both the talkers using the voice signals of the talkers.

  The stress measuring device according to the present invention can measure the stress of both parties using voice signals of both of the speaking parties. For example, this stress measuring device can be installed as a function of a mobile phone. When making a call on a mobile phone, it is possible to determine whether the other party on the mobile phone is under stress.

DESCRIPTION OF SYMBOLS 1 Stress measuring device 2 Voice A input part 3 Voice A buffer 4 Voice A analysis part 5 Voice A voice unit extraction part 6 Voice A frequency analysis part 7 Voice A amplitude analysis part 8 Voice A data storage part 9 Voice B input part 10 Voice B buffer 11 Audio B analysis unit 12 Audio B audio unit extraction unit 13 Audio B frequency analysis unit 14 Audio B amplitude analysis unit 15 Audio B data storage unit 16 Display unit 17 Audio A stress determination unit 18 Audio B stress determination unit 19 Audio A Stress parameter X _An calculation unit 20 Audio A stress parameter Y _An calculation unit 21 Audio B stress parameter X _Bn calculation unit 22 Audio B stress parameter Y _Bn calculation unit 50 Speaker 51 Liquid crystal display 53 Electrode 54 Power supply

Claims (9)

A voice input unit that inputs voices of both speakers during the conversation and generates voice signals;
A voice analysis unit that calculates a stress index from the frequency and amplitude of a voice signal from the voice input unit;
A voice stress determination unit that determines a stress value of a speaker based on the stress index;
Stress measuring device with The voice analysis unit
A voice unit extraction unit that divides the voice signal into a voice part and a voiceless part and extracts only the voice part as a voice unit;
An audio frequency analyzer that calculates an average frequency of the audio unit and calculates a frequency difference value from the average frequency of the audio unit preceding the audio unit;
A first stress variable calculation unit having a quotient of the frequency difference value and the average frequency as a first stress variable;
An audio amplitude analyzer that calculates an average amplitude of the audio unit and calculates an amplitude difference value from the average amplitude of the preceding audio unit;
The stress measurement apparatus according to claim 1, further comprising a second stress variable calculation unit that uses a quotient of the amplitude difference value and the average frequency as a second stress variable. The voice unit extraction unit
The stress measuring device according to claim 1, wherein a voiced portion between two adjacent voiceless parts is used as a voiced part if no voice is present in the voice signal for a predetermined time or longer. The first stress variable calculator is
A frequency difference △ f _An between the frequency f _An-1 in the frequency f _An and the previous speech unit in the speech unit calculated by (Equation 1),
The using the equation (2) and the frequency f _i in the frequency difference △ f _An and any speech unit 1
The stress measurement device according to claim 2, wherein the stress parameter X _An is calculated.

Here, N is the number of audio units. The second stress variable calculation unit includes:
Amplitude difference value between the amplitude R _An-1 in the amplitude R _An and the previous speech unit in the speech unit △
_Calculate R _An by (Equation 3)
3. The stress measuring device according to claim 2, wherein the second stress parameter Y _An is calculated using the amplitude difference value ΔR _An and the amplitude Ri in an arbitrary voice unit using (Expression 4).

Here, N is the number of audio units. The voice stress determination unit
The stress measurement device according to claim 2, wherein an average value of the first and second stress variables is set as a third stress index, and if the third stress index is positive, it is determined as stress. The third stress index is
The stress measuring device according to claim 6, wherein the first stress parameter X _An and the second stress parameter Y _An are calculated using the number (5).

Here, Z _A is the third stress index, and N is the number of voice units. The third stress index is
The stress measuring apparatus according to claim 6, wherein the two stress parameters X _An and Y _An to which weights are added are calculated using the number (6).

Here, Z _A is a third stress index, N is the number of voice units, a is a coefficient indicating the weight of the stress parameter X _An , and b is a coefficient indicating the weight of the stress parameter Y _An . The stress measurement apparatus according to claim 1, further comprising a display unit for outputting a speaker stress value determined by the voice stress determination unit.

Images