Granted Patent B2

US 12,217,767 · App. 17/630,090 · Granted Feb 4, 2025

Inventors: Fumiya Nagashima (Kanagawa, JP); Kengo Akimoto (Kanagawa, JP); Tatsuya Okano (Kanagawa, JP); Yusuke Koumura (Kanagawa, JP); Seiko Inoue (Kanagawa, JP)

Assignee: Semiconductor Energy Laboratory Co., Ltd.

G10L21/0272G10L25/30

Abstract

Claims (37)

1. An audio device comprising:

a sound sensor portion;

a sound separation portion;

a sound determination portion;

a processing portion; and

an output portion,

wherein the audio device is an earphone or a headphone,

wherein the sound sensor portion is configured to sense first sound,

wherein the sound determination portion is configured to store a feature quantity of a voice of a user,

wherein the sound determination portion is configured to determine, with a machine learning model, whether the first sound has the stored feature quantity,

wherein the processing portion is configured to analyze an instruction contained in the first sound and, when the feature quantity of the first sound is the stored feature quantity, to generate a signal representing content of the instruction, and, when the feature quantity of the first sound is not the stored feature quantity, to not generate the signal representing content of the instruction,

wherein second sound is sound to cancel noise in the first sound,

wherein third sound is sound electrically input to the audio device,

wherein fourth sound is synthesized sound of the second sound and the third sound,

wherein the processing portion is configured to generate the second sound and the third sound, and

wherein the output portion is configured to output the fourth sound.

2. The audio device according to claim 1 ,

wherein learning for the machine learning model is performed using supervised learning in which a voice is learning data and a label indicating whether the storing is to be performed is training data.

3. The audio device according to claim 1 ,

wherein the machine learning model is a neural network model.

4. The audio device according to claim 1 ,

wherein the second sound is sound having a phase opposite to a phase of the noise in the first sound.

5. An operation method of an audio device, comprising:

sensing first sound;

determining, with a machine learning model, whether a feature quantity of the first sound is a stored feature quantity of a voice of a user;

analyzing an instruction contained in the first sound;

generating a signal representing content of the instruction when the feature quantity of the first sound is the stored feature quantity and not generating the signal representing content of the instruction when the feature quantity of the first sound is not the stored feature quantity;

generating second sound to cancel noise in the first sound;

electrically inputting third sound to the audio device;

generating fourth sound by synthesizing the second sound and the third sound; and

outputting the fourth sound.

6. The operation method of the audio device, according to claim 5 ,

wherein learning for the machine learning model is performed using supervised learning in which a voice is used as learning data and a label indicating whether storing is to be performed is used as training data.

7. The operation method of the audio device, according to claim 5 ,

wherein the machine learning model is a neural network model.

8. The operation method of the audio device, according to claim 5 ,

wherein the second sound is sound having a phase opposite to a phase of the noise in the first sound.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (29)