# US 2019/0378534 A1 · Apparatus for Generating Relations Between Feature Amounts of Audio and Scene Types and Method Therefor

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2019/0378534 A1
- Document ID: US20190378534A1
- Status: Published application
- Kind: A1
- Application number: 16/550,404
- Filed: 2019-08-26
- Published: 2019-12-12
- Earliest claimed: 2017-02-27
- Assignees: Yamaha Corporation
- Applicants: Yamaha Corporation
- Inventors: Yuta YUYAMA; Keita Arimoto
- Web page: https://patentgrep.com/patent/US20190378534A1

## CPC

- G06F 16/75: Clustering; Classification
- G10L 25/57: for processing of video signals
- G06F 16/61: Indexing; Data structures therefor; Storage structures
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 20/00: Machine learning
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G10L 25/27: characterised by the analysis technique
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

An apparatus for generating relations between feature amounts of audio and scene types includes at least one processor and a memory. Upon execution of the instructions the processor is caused to set one of the scene types in accordance with an instruction from a user to indicate one of clusters classifying the feature amounts of audio in one or more pieces of content. The processor is also caused to generate a plurality of pieces of learning data, each representative of a feature amount, from among the feature amounts, that belongs to the cluster and the scene type set for the cluster. The processor is also caused to generate an identification model representative of relations between the feature amounts of audio and the scene types by performing machine learning using the plurality of pieces of learning data.

## Claims

### Claim 1

An apparatus for generating relations between feature amounts of audio and scene types, comprising:
at least one processor; and
a memory that is operatively coupled to the at least one processor, the memory storing instructions executable by the processor, wherein upon execution of the instructions the processor is caused to:
set one of the scene types in accordance with an instruction from a user to indicate one of clusters classifying the feature amounts of audio in one or more pieces of content;
generate a plurality of pieces of learning data, each representative of a feature amount, from among the feature amounts, that belongs to the cluster and the scene type set for the cluster; and
generate an identification model representative of relations between the feature amounts of audio and the scene types by performing machine learning using the plurality of pieces of learning data.

### Claim 2 (depends on 1)

The apparatus according to claim 1, wherein the identification model is a statistical model identifying one scene type from a feature amount of audio in content.

### Claim 3 (depends on 1)

The apparatus according to claim 1, wherein the identification model is a statistical model that identifies a likelihood relating to each of the scene types from a feature amount of audio in content.

### Claim 4 (depends on 1)

The apparatus according to claim 1, wherein the scene type is a classification of a scene represented by the content.

### Claim 5

A method for generating relations between feature amounts of audio and scene types comprising:
setting one of the scene types in accordance with an instruction from a user to indicate one of clusters classifying the feature amounts of audio in one or more pieces of content; and
generating a plurality of pieces of learning data, each representative of a feature amount, from among the feature amounts, that belongs to the cluster and the scene type set for the cluster.

### Claim 6 (depends on 5)

The method according to claim 5, further comprising:
generating an identification model representative of relations between the feature amounts of audio and the scene types by performing machine learning using the plurality of pieces of learning data.

### Claim 7 (depends on 6)

The method according to claim 6, wherein the identification model is a statistical model identifying one scene type from a feature amount of audio in content.

### Claim 8 (depends on 6)

The method according to claim 6, wherein the identification model is a statistical model that identifies a likelihood relating to each of the scene types from a feature amount of audio in content.

### Claim 9 (depends on 5)

The method according to claim 5, wherein the scene type is a classification of a scene represented by the content.