Utterance state detection device and utterance state detection method

• US 20110282666A1
• Filed: 04/21/2011
• Published: 11/17/2011
• Est. Priority Date: 04/22/2010
• Status: Active Grant

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Kamano, Akira, Matsuo, Naoshi, Harada, Shouji, Washio, Nobuyuki
• Granted Patent: US 9,099,088 B2
• Time in Patent Office: Days
• US Class Current: 704/246
• CPC Class Codes: G10L 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices
• CPC Class Codes: G10L 25/48 specially adapted for particular use

Abstract

An utterance state detection device includes an user voice stream data input unit that gets user voice stream data of an user, a frequency element extraction unit that extracts high frequency elements by frequency-analyzing the user voice stream data, a fluctuation degree calculation unit that calculates a fluctuation degree of the high frequency elements thus extracted every unit time, a statistic calculation unit that calculates a statistic every certain interval based on a plurality of the fluctuation degrees in a certain period of time, and an utterance state detection unit that detects an utterance state of a specified user based on the statistic obtained from user voice stream data of the specified user.

Claims

1. An utterance state detection device, comprising:

an user voice stream data input unit that acquires user voice stream data of an user;

a frequency element extraction unit that extracts high frequency elements by frequency-analyzing the user voice stream data;

a fluctuation degree calculation unit that calculates a fluctuation degree of the extracted high frequency elements every unit time;

a statistic calculation unit that calculates a statistic every certain interval based on a plurality of the fluctuation degrees in a certain period of time; and

an utterance state detection unit that detects an utterance state of a specified user based the statistic obtained from user voice stream data of the specified user.

2. The utterance state detection device according to claim 1, wherein the utterance state detection unit detects that the utterance state of the specified user is whether in an ordinary state or in an unordinary state by using an utterance state detection parameter calculated based on the statistic obtained from user voice stream data generated when an unspecified user is making an utterance in a known utterance state.

3. The utterance state detection device according to claim 2, further comprising an utterance state detection parameter generation unit that produces the utterance state detection parameter based on the statistic obtained from user voice stream data generated when an unspecified user is making an utterance in a known utterance state.

4. The utterance state detection device according to claim 1, wherein the high frequency elements are extracted so as to include no first formant.

5. The utterance state detection device according to claim 1, wherein the high frequency elements have a frequency of 2 kHz or higher.

6. The utterance state detection device according to claim 1, wherein the fluctuation degree calculation unit calculates the fluctuation degree based on a formula:

fluctuation degree=log(a geometric mean of power spectra of high frequency elements)/log(an arithmetic mean of power spectrum of high frequency elements).

7. The utterance state detection device according to claim 1, wherein the fluctuation degree calculation unit calculates the fluctuation degree based on a formula:

fluctuation degree=log(a geometric mean of residual power spectra of high frequency elements)/log(an arithmetic mean of residual power spectrum of high frequency elements).

8. The utterance state detection device according to claim 1, wherein the fluctuation degree calculation unit calculates the fluctuation degree based on a formula:

fluctuation degree=a geometric mean of power spectra of high frequency elements/an arithmetic mean of power spectrum of high frequency elements.

9. The utterance state detection device according to claim 1, wherein the fluctuation degree calculation unit calculates the fluctuation degree based on a variance of residual power spectra of high frequency elements.

10. The utterance state detection device according to claim 1, wherein the fluctuation degree calculation unit calculates the fluctuation degree based on a quartile range of residual power spectra of high frequency elements.

11. The utterance state detection device according to claim 1, further comprising a signal-to-noise (SN) ratio calculation unit that calculates an SN ratio of the user voice stream data, wherein

the utterance state detection unit detects an utterance state when the SN ratio is larger than a threshold in the certain period of time in which the statistic is calculated.

14. The utterance state detection device according to claim 1, further comprising:

a reply model storage unit that stores therein a reply model including statistically processed information relating to a reply of the user in a normal state;

an interval detection unit that estimates a reply interval with a voice interval shorter than a certain threshold in the user voice stream data; and

a state determination unit that determines an ordinary degree of the reply interval with reference to a reply model stored in the reply model storage unit when the interval detection unit estimates that the reply interval continuously appears.

15. The utterance state detection device according to claim 14, wherein the state determination unit determines an ordinary degree of the reply interval by further using any one or more than one of the number of reply interval in the voice interval, a length of the reply interval in an overall length of the voice interval, a length between replies in the reply interval, and a total length of the reply interval.

16. The utterance state detection device according to claim 14, further comprising:

a call duration calculation unit that calculates a call duration of a call made by the user; and

an utterance rate calculation unit that calculates an utterance rate of the user based on a summation of interval lengths of voice intervals in the call and a summation of syllable series in the call, wherein

the state determination unit determines an ordinary degree of the reply interval by using any one or more than one of a ratio of the summation of interval lengths of voice intervals to the call duration calculated by the call duration calculation unit, a ratio of a summation of interval lengths of the reply interval to the call duration, and the utterance rate calculated by the utterance rate calculation unit.

17. The utterance state detection device according to claim 1, further comprising:

a call duration calculation unit that calculates a call duration of a call made by the user; and

an utterance rate calculation unit that calculates an utterance rate of the user based on a summation of interval lengths of voice intervals in the call and a summation of syllable series in the call; and

a total determination unit that comprehensively determines the utterance state by using a result of the utterance state detected by the utterance state detection unit and an utterance rate calculated by the utterance rate calculation unit.

12. A computer readable storage medium containing an utterance state detection program for detecting an utterance state of an user that, the utterance state detection program causing a computer to perform:

acquiring user voice stream data of the user;

extracting high frequency elements by frequency-analyzing the user voice stream data;

calculating a fluctuation degree of the extracted high frequency elements every unit time;

calculating a statistic every certain interval based on a plurality of the fluctuation degrees in a certain period of time; and

detecting an utterance state of a specified user based the statistic obtained from user voice stream data of the specified user.

13. An utterance state detection method, comprising:

acquiring user voice stream data of an user;

extracting high frequency elements by frequency-analyzing the user voice stream data;

calculating a fluctuation degree of the extracted high frequency elements every unit time;

calculating a statistic every certain interval based on a plurality of the fluctuation degrees in a certain period of time; and

detecting an utterance state of a specified user based the statistic obtained from user voice stream data of the specified user.