METHOD AND SYSTEM FOR LAUGHTER DETECTION

• US 20080195385A1
• Filed: 02/11/2007
• Published: 08/14/2008
• Est. Priority Date: 02/11/2007
• Status: Active Grant

• Current Assignee: Nice Ltd
• Original Assignee: Nice Systems Limited (Nice Ltd)
• Inventors: PEREG, Oren, WASSERBLAT, Moshe
• Granted Patent: US 8,571,853 B2
• Time in Patent Office: Days
• US Class Current: 704/231
• CPC Class Codes: G10L 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices

Abstract

A method and apparatus for laughter detection. Laughter is detected through the presence of a sequence of at least a predetermined number such as three consecutive bursts, each burst comprising a voiced portion and an unvoiced portion. After detecting bursts, n-tuples such as triplets are detected, and a likelihood of each burst N-tuple to represent laughter is provided by comparison to predetermined thresholds. Finally, a total score is assigned to the signal based on the grades associated with the triplets and parameters such as the distance between the N-tuples, the total score representing the probability that the audio signal comprises a laughter episode. The method and apparatus preferably comprise a training step and module for determining the thresholds according to manually marked audio signals.

Claims

1. A method for detecting an at least one laughter episode in an audio signal, the method comprising the steps of:

detecting an at least one burst in the audio signal;

detecting at least predetermined number of consecutive bursts;

extracting an at least one feature from the at least predetermined number of consecutive bursts;

determining whether the at least one feature complies with an at least one predetermined threshold; and

if the at least one feature complies with the at least one predetermined threshold, outputting the at least predetermined number of consecutive bursts as a laughter episode.

2. The method of claim 1 wherein the at least one burst comprises a voiced segment followed by an unvoiced segment.

3. The method of claim 1 wherein the at least one burst comprises a voiced segment followed by a silent segment.

4. The method of claim 1 further comprising a preprocessing step for enhancing the audio signal for further processing.

5. The method of claim 1 wherein the at least one threshold is associated with an at least one score, said at least one score assigned to the at least predetermined number of consecutive bursts.

6. The method of claim 1 wherein the detection of the at least one burst comprises the steps of:

dividing the audio signal into frames;

detecting for each of the frames whether it is voiced, unvoiced, or silent;

determining an at least one voiced section comprising voiced frames, and an at least one unvoiced section comprising unvoiced frames or an at least one silent section comprising silent frames; and

detecting an at least one burst as one of the at least one voiced sections followed by one of the at least one unvoiced sections or one of the at least one silent sections.

7. The method of claim 6 wherein each of the frames comprises between about 20 samples and about 2000 samples.

8. The method of claim 1 further comprising a training step for determining the at least one predetermined threshold.

9. The method of claim 8 wherein the training step comprises the steps of:

detecting an at least one burst in the audio signal;

detecting at least predetermined number of consecutive bursts, associated with an at least one part of the audio signal;

extracting an at least one feature from the at least one part of the audio signal;

receiving an at least one manually marked episode; and

determining an at least one threshold, so that the at least one feature complies with the at least one threshold, if the at least one part of the audio signal at least partly overlaps with the at least one manually marked episode.

10. The method of claim 9 further comprising a performance measurement step for measuring the performance of the method associated with the at least one threshold.

11. The method of claim 1 wherein the at least one feature is any of the group consisting of:

mean voiced length;

mean duty cycle;

mean energy ratio;

Mean distance between voiced sections;

or an itakura distance.

12. The method of claim 1 further comprising a step of capturing the audio signal.

13. The method of claim 1 further comprising a step of interaction analysis of the audio signal.

14. The method of claim 1 further comprising a step of quality evaluation for the audio signal.

15. The method of claim 14 wherein the quality evaluation step relates to an entity selected from the group consisting of:

agent evaluation;

interaction evaluation;

product evaluation;

sale evaluation;

or service evaluation.

16. The method of claim 1 when performed as part of an agent training.

17. The method of claim 1 wherein the audio signal is captured from any of the group consisting of:

a telephone;

a voice over IP device;

a microphone;

or an audio part of a video capturing.

18. An apparatus for detecting an at least one laughter episode in an audio signal, the apparatus comprising:

a burst detection component for detecting an at least one burst in the audio signal;

a burst N-tuple detection component for detecting at least predetermined number of consecutive bursts;

a burst N-tuple feature extraction component for extracting an at least one feature from the at least predetermined number of consecutive bursts;

a scoring component for scoring the least one feature against an at least one predetermined threshold; and

a decision component for determining whether the at least one feature is associated with a laughter episode, according to whether the at least one feature complies with the at least one predetermined threshold.

19. The apparatus of claim 18 further comprising a total scoring component for assigning a total score to the audio signal.

20. The apparatus of claim 18 further comprising:

a threshold set adjustment component for determining the at least one predetermined threshold; and

a performance measurement component for evaluating the at least one threshold.

21. The apparatus of claim 20 further comprising a manual marking component for marking an at least one laughter episode in an audio signal.

22. The apparatus of claim 18 further comprising a quality monitoring component for receiving a laughter episode indication for quality monitoring purposes.

23. The apparatus of claim 22 wherein the monitoring purposes are selected from the group consisting of:

agent monitoring;

interaction monitoring;

product monitoring;

sale monitoring, or service monitoring.

24. The apparatus of claim 18 farther comprising an interaction analysis component for receiving a laughter episode indication and analyzing an interaction associated with the audio signal.

25. A computer readable storage medium containing a set of instructions for a general purpose computer, the set of instructions comprising:

detecting an at least one burst in an audio signal;

detecting an at least predetermined number of consecutive bursts;

extracting an at least one feature from the at least predetermined number of consecutive bursts;

determining whether the at least one feature complies with an at least one predetermined threshold; and

if the at least one feature complies with the at least one predetermined threshold, outputting the at least predetermined number of consecutive bursts as a laughter episode.