Sound Event Detecting Module and Method Thereof

• US 20120143610A1
• Filed: 12/30/2010
• Published: 06/07/2012
• Est. Priority Date: 12/03/2010
• Status: Active Grant

• Current Assignee: Industrial Technology Research Institute
• Original Assignee: Industrial Technology Research Institute
• Inventors: Wang, Yuh-Ching, Li, Kuo-Yuan
• Granted Patent: US 8,655,655 B2
• Time in Patent Office: Days
• US Class Current: 704/256
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/10 using distance or distortion measures between unknown speech and reference templates
• 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

A sound event detecting module for detecting whether a sound event with characteristic of repeating is generated. A sound end recognizing unit recognizes ends of sounds according to a sound signal to generate sound sections and multiple sets of feature vectors of the sound sections correspondingly. A storage unit stores at least M sets of feature vectors. A similarity comparing unit compares the at least M sets of feature vectors with each other, and correspondingly generates a similarity score matrix, which stores similarity scores of any two of the sound sections of the at least M of the sound sections. A correlation arbitrating unit determines the number of sound sections with high correlations to each other according to the similarity score matrix. When the number is greater than one threshold value, the correlation arbitrating unit indicates that the sound event with the characteristic of repeating is generated.

Claims

1. A sound event detecting module for detecting whether a sound event with the characteristic of repeating is generated, the module comprising:

a sound end recognizing unit for recognizing ends of sounds according to a sound signal to generate a plurality of sound sections and a plurality of sets of feature vectors of the sound sections correspondingly;

a storage unit for storing at least M sets of feature vectors of the sets of feature vectors, wherein M is a positive integer;

a similarity comparing unit for comparing the at least M sets of feature vectors with each other, and generating a similarity score matrix correspondingly, wherein the similarity score matrix stores similarity scores of any two of the sound sections; and

a correlation arbitrating unit for determining the number of sound sections having high correlations to each other according to the similarity score matrix, wherein when the number is greater than a threshold value, it is indicated that the sound event with the characteristic of repeating is generated.

2. The sound event detecting module according to claim 1, wherein the sound event with the characteristic of repeating comprises at least one of a sound of help, a sound of cry, a sound of groan, a sound of cough and a sound of asthma.

3. The sound event detecting module according to claim 1, wherein the similarity comparing unit uses at least one of a sound event template, a sound event statistics model, a sound event possibility model, a hidden Markov model and a neural network to compare the sets of feature vectors with each other.

4. The sound event detecting module according to claim 1, further comprising a non-target sound filtering unit for determining whether to ignore the sound event with the characteristic of repeating according to the sets of feature vectors, wherein each of the sets of feature vectors comprises at least one of a sound section length, a zero crossing rate, energy of a specific frequency range and a frequency range of main energy.

5. The sound event detecting module according to claim 1, further comprising an event classifying unit for classifying the sound event with the characteristic of repeating according to the sets of feature vectors corresponding to the sound event with the characteristic of repeating wherein each of the sets of feature vectors comprises at least one of a Mel-scale frequency cepstral coefficient, a cepstrum, a linear prediction coefficient, a linear spectral pair, energy of a specific frequency range, a frequency range of main energy, a high/low frequency energy ratio, a sound section length, a zero crossing rate and sound energy.

6. The sound event detecting module according to claim 1, further comprising an event reporting unit for executing at least one of a reporting function, a warning function and a reminding function when the sound event with the characteristic of repeating is generated.

7. The sound event detecting module according to claim 1, wherein the sound sections are continuous sound sections, the at least M sets of feature vectors are feature vectors of at least M sound sections latest occurred, and the correlation arbitrating unit further outputs the sound event with the characteristic of repeating.

8. The sound event detecting module according to claim 1, wherein the storage unit is implemented by a ring buffer, a first-in-first-out (FIFO) queue buffer or a PingPong buffer.

9. A sound event detection method for detecting whether a sound event with the characteristic of repeating is generated, the method comprising the steps of:

recognizing ends of sounds according to a sound signal to generate a plurality of sound sections and a plurality of sets of feature vectors of the sound sections correspondingly;

storing at least M sets of feature vectors of the sets of feature vectors, wherein M is a positive integer;

comparing the at least M sets of feature vectors with each other and generating a similarity score matrix correspondingly, wherein the similarity score matrix stores similarity scores of any two of the sound sections; and

determining the number of sound sections having high correlations to each other according to the similarity score matrix, wherein when the number is greater than a threshold value, it is indicated that the sound event with the characteristic of repeating is generated.

10. The method according to claim 9, wherein the sound event with the characteristic of repeating comprises at least one of a sound of help, a sound of cry, a sound of groan, a sound of cough and a sound of asthma.

11. The method according to claim 9, wherein the step of comparing uses at least one of a sound event template, a sound event statistics model, a sound event possibility model, a hidden Markov model and a neural network.

12. The method according to claim 9, further comprising the step of:

determines whether to ignore the sound event with the characteristic of repeating according to the sets of feature vectors, wherein each of the sets of feature vectors comprises at least one of a sound section length, a zero crossing rate, energy of a specific frequency range and a frequency range of main energy.

13. The method according to claim 9, further comprising the step of:

classifying the sound event with the characteristic of repeating according to the sets of feature vectors corresponding to the sound event with the characteristic of repeating wherein each of the sets of feature vectors comprises at least one of a Mel-scale frequency cepstral coefficient, a cepstrum, a linear prediction coefficient, a linear spectral pair, energy of a specific frequency range, a frequency range of main energy, a high/low frequency energy ratio, a sound section length, a zero crossing rate and sound energy.

14. The method according to claim 9, further comprising the step of:

executing at least one of a reporting function, a warning function and a reminding function when the sound event with the characteristic of repeating is generated.

15. The method according to claim 9, wherein the sound sections are continuous sound sections, the at least M sets of feature vectors are feature vectors of at least M sound sections latest occurred, and the method further comprises the step of:

outputting the sound event with the characteristic of repeating.

16. The method according to claim 9, wherein the at least M sets of feature vectors are stored in a ring buffer, a FIFO queue buffer or a PingPong buffer.