Speech processing apparatus, speech processing method and computer-readable medium

1. A speech processing apparatus, comprising:

a memory storing instructions; and

at least one processor configured to process the instructions to perform steps including:

calculating distribution of types of sounds included in a speech signal by using a speech model configured to identify the types of sounds represented by the speech signal, wherein the types of sounds represent sound's classes clustered based on similarity of the sounds;

calculating an acoustic diversity degree representing a degree of appearance of each of the types of sounds included in the speech signal;

calculating a recognition feature to recognize an attribute of the speaker;

compensating for the recognition feature so as to have a smaller bias of the types of sounds in the speech signal, by using the acoustic diversity degree value;

generating a recognition result based on the compensated recognition feature; and

transmitting the recognition result.

2. The speech processing apparatus according to claim 1 , wherein the types of sounds are types of phonemes.

3. The speech processing apparatus according to claim 1 , wherein the at least one processor is further configured to process the instructions to perform steps including:

calculating the degree of appearance of each of the types of sounds by using the speech model; and

calculating the acoustic diversity degree value by using the degree of appearance;

wherein the degree of appearance represents distribution of the types of sounds included in the speech signal.

4. The speech processing apparatus according to claim 1 , wherein the at least one processor is further configured to process the instructions to perform steps including:

calculating the compensated recognition feature of the speech signal based on an acoustic diversity degree, a recognition feature of each of at least one first speech signal, and a recognition feature of each of at least one second speech signal.

5. The speech processing apparatus according to claim 4 , wherein the first speech signal includes one or more speech signals obtained by segmenting the at least one second speech signal.

6. The speech processing apparatus according to claim 4 , wherein the at least one processor is further configured to process the instructions to perform steps including:

compensating for the recognition feature of the speech signal by using a neural network.

7. The speech processing apparatus according to claim 1 , wherein the at least one processor is further configured to process the instructions to perform steps including:

recognizing the attribute of the speaker based on the speech signal,

wherein the attribute represents at least one of information indicating the speaker, a language spoken by the speaker, an emotional expression included in the speech signal, and a character of a speaker wherein the character is assumed from the speech signal.

8. A speech processing method comprising:

calculating, using at least one processor, distribution of types of sounds included in a speech signal by using a speech model configured to identify the types of sounds represented by the speech signal, wherein the types of sounds represent sound's classes clustered based on similarity of the sounds;

calculating, using the at least one processor, an acoustic diversity degree representing a degree of appearance of each of the types of sounds included in the speech signal;

calculating a recognition feature to recognize an attribute of the speaker;

compensating for the recognition feature so as to have a smaller bias of the types of sounds in the speech signal, using the at least one processor, by using the acoustic diversity degree value;

generating a recognition result based on the compensated recognition feature; and

transmitting the recognition result.

9. The speech processing method of claim 8 , wherein the types of sounds are types of phonemes.

10. The speech processing method of claim 8 , further comprising:

calculating the degree of appearance of each of the types of sounds by using the speech model; and

calculating the acoustic diversity degree value by using the degree of appearance;

wherein the degree of appearance represents distribution of the types of sounds included in the speech signal.

11. The speech processing method of claim 8 , further comprising:

calculating the compensated recognition feature of the speech signal based on an acoustic diversity degree, a recognition feature of each of at least one first speech signal, and a recognition feature of each of at least one second speech signal.

12. The speech processing method of claim 11 , wherein the first speech signal includes one or more speech signals obtained by segmenting the at least one second speech signal.

13. The speech processing method of claim 11 , further comprising:

compensating for the recognition feature of the speech signal by using a neural network.

14. The speech processing method of claim 8 , further comprising:

recognizing the attribute of the speaker based on the speech signal,

wherein the attribute represents at least one of information indicating the speaker, a language spoken by the speaker, an emotional expression included in the speech signal, and a character of a speaker wherein the character is assumed from the speech signal.

15. A non-transitory computer-readable storage medium storing instructions that when executed by a computer enable the computer to implement a method comprising:

calculating, using at least one processor, distribution of types of sounds included in a speech signal using a speech model configured to identify the types of sounds represented by the speech signal, wherein the types of sounds represent sound's classes clustered based on similarity of the sounds;

calculating, using the at least one processor, an acoustic diversity degree representing a degree of appearance of each of the types of sounds included in the speech signal;

calculating a recognition feature to recognize an attribute of the speaker;

compensating for the recognition feature so as to have a smaller bias of the types of sounds in the speech signal, using the at least one processor, by using the acoustic diversity degree value;

generating a recognition result based on the compensated recognition feature; and

transmitting the recognition result.

16. The non-transitory computer-readable storage medium according to claim 15 , wherein the types of sounds are types of phonemes.

17. The non-transitory computer-readable storage medium according to claim 15 , wherein the method further comprises:

calculating the degree of appearance of each of the types of sounds by using the speech model; and

calculating the acoustic diversity degree value by using the degree of appearance;

wherein the degree of appearance represents distribution of the types of sounds included in the speech signal.

18. The non-transitory computer-readable storage medium according to claim 15 , wherein the method further comprises:

calculating the compensated recognition feature of the speech signal based on an acoustic diversity degree, a recognition feature of each of at least one first speech signal, and a recognition feature of each of at least one second speech signal.

19. The non-transitory computer-readable storage medium according to claim 18 , wherein the first speech signal includes one or more speech signals obtained by segmenting the at least one second speech signal.

20. The non-transitory computer-readable storage medium according to claim 18 , wherein the method further comprises:

compensating for the recognition feature of the speech signal by using a neural network.

21. The non-transitory computer-readable storage medium according to claim 15 , wherein the method further comprises:

recognizing the attribute of the speaker based on the speech signal,

wherein the attribute represents at least one of information indicating the speaker, a language spoken by the speaker, an emotional expression included in the speech signal, and a character of a speaker wherein the character is assumed from the speech signal.