# US 2020/0286486 A1 · Voice identification method, device, apparatus, and storage medium

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

- Number: US 2020/0286486 A1
- Document ID: US20200286486A1
- Status: Published application
- Kind: A1
- Application number: 16/801,742
- Filed: 2020-02-26
- Published: 2020-09-10
- Earliest claimed: 2019-03-05
- Assignees: Baidu Online Network Technology (Beijing) Co., Ltd.
- Applicants: Baidu Online Network Technology (Beijing) Co., Ltd
- Inventors: Ye SONG; Long ZHANG; Pengpeng JIE
- Web page: https://patentgrep.com/patent/US20200286486A1

## CPC

- G10L 15/10: using distance or distortion measures between unknown speech and reference templates
- G10L 15/26: Speech to text systems (G10L15/08 takes precedence)
- G10L 15/02: Feature extraction for speech recognition; Selection of recognition unit
- G10L 15/05: Word boundary detection
- G10L 15/06: Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice (G10L15/14 takes precedence)
- G10L 15/1807: {using prosody or stress}
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

A voice identification method, device, apparatus, and a storage medium are provided. The method includes: receiving voice data; and performing a voice identification on the voice data, to obtain first text data associated with the voice data; determining common text data in a preset fixed data table, wherein a similarity between a pronunciation of the determined common text data and a pronunciation of the first text data meets a preset condition, wherein the determined common text data is a voice identification result with an occurrence number larger than a first preset threshold; and replacing the first text data with the determined common text data. Voice identification accuracy is improved.

## Claims

### Claim 1

A voice identification method, comprising:
receiving voice data; and performing a voice identification on the voice data, to obtain first text data associated with the voice data;
determining common text data in a preset fixed data table, wherein a similarity between a pronunciation of the determined common text data and a pronunciation of the first text data meets a preset condition, wherein the determined common text data is a voice identification result with an occurrence number larger than a first preset threshold; and
replacing the first text data with the determined common text data.

### Claim 2 (depends on 1)

The voice identification method according to claim 1, wherein the performing a voice identification on the voice data comprises: performing an offline voice identification on the voice data, wherein
the common text data is obtained by an online voice identification and has the occurrence number larger than the first preset threshold.

### Claim 3 (depends on 1)

The voice identification method according to claim 1, further comprising: establishing the preset fixed data table by:
performing an online voice identification on a pre-obtained voice data, to obtain second text data associated with the pre-obtained voice data;
determining a semantic type of the second text data is offline processable;
performing a word segment on the second text data, to obtain at least one word; and
determine whether the obtained at least one word or the second text data is in a fixed data table; updating an occurrence number of the at least one word or of the second text data in a temporary data table, in a case where the obtained at least one word or the second text data is not comprised in the fixed datable; and recording a word or second text data having an occurrence number larger than a second preset threshold into the fixed data table, as a candidate common text data in the fixed data table.

### Claim 4 (depends on 3)

The voice identification method of claim 3, further comprising:
releasing a storage space of the temporary data table in a Least Recently Used mode, in a case where a storage amount of the temporary data table reaches a storage upper limit.

### Claim 5 (depends on 1)

The voice identification method according to claim 1, wherein semantics parsing result of the common text data is stored in the preset fixed data table; and
after the determining common text data in a preset fixed data table, the method further comprises extracting the parsed semantics of the at common text data from the preset fixed data table.

### Claim 6 (depends on 1)

The voice identification method according to claim 1, wherein the determining common text data in a preset fixed data table comprises:
comparing the first text data with each candidate common text data in the preset fixed data table, by comparing phonemes at the same position of the first text data and the candidate common text data one by one to determine a same phoneme; and
determining that the similarity between a pronunciation of the determined common text data and a pronunciation of the first text data meets the preset condition, in a case where a ratio of the number of same phonemes to the number of all phonemes is larger than a preset ratio threshold.

### Claim 7

A voice identification device, comprising:
one or more processors; and
a memory for storing one or more programs, wherein
the one or more programs are executed by the one or more processors to enable the one or more processors to:
receive voice data; and perform a voice identification on the voice data, to obtain first text data associated with the voice data;
determine common text data in a preset fixed data table, wherein a similarity between a pronunciation of the determined common text data and a pronunciation of the first text data meets a preset condition, wherein the determined common text data is a voice identification result with an occurrence number larger than a first preset threshold; and
replace the first text data with the determined common text data, in response to a determination of the common text data in the preset fixed data table.

### Claim 8 (depends on 7)

The voice identification device according to claim 7, wherein the one or more programs are executed by the one or more processors to enable the one or more processors to perform an offline voice identification on the voice data; and
set the fixed data table, wherein the fixed data table comprises one or more common text data, and the common text data is obtained by an online voice identification and has the occurrence number larger than the first preset threshold.

### Claim 9 (depends on 8)

The voice identification device according to claim 8, wherein the one or more programs are executed by the one or more processors to enable the one or more processors to:
perform an online voice identification on a pre-obtained voice data, to obtain second text data associated with the pre-obtained voice data;
determine a semantic type of the second text data is offline processable;
perform a word segment on the second text data, to obtain at least one word; and
determine whether the obtained at least one word or the second text data is in a fixed data table; update an occurrence number of the at least one word or of the second text data in a temporary data table, in a case where the obtained at least one word or the second text data is not comprised in the fixed datable; and record a word or second text data having an occurrence number larger than a second preset threshold into the fixed data table, as a candidate common text data in the fixed data table.

### Claim 10 (depends on 9)

The voice identification device of claim 9, wherein the one or more programs are executed by the one or more processors to enable the one or more processors to:
release a storage space of the temporary data table in a Least Recently Used mode, in a case where a storage amount of the temporary data table reaches a storage upper limit.

### Claim 11 (depends on 7)

The voice identification device according to claim 7, wherein semantics parsing result of the common text data is stored in the preset fixed data table; and
the one or more programs are executed by the one or more processors to enable the one or more processors to extract the parsed semantics of the at common text data from the preset fixed data table, in response to a determination of the common text data in the preset fixed data table.

### Claim 12 (depends on 7)

The voice identification device according to claim 7, wherein the one or more programs are executed by the one or more processors to enable the one or more processors to:
compare the first text data with each candidate common text data in the preset fixed data table, by comparing phonemes at the same position of the first text data and the candidate common text data one by one to determine a same phoneme; and
determine that the similarity between a pronunciation of the determined common text data and a pronunciation of the first text data meets the preset condition, in a case where a ratio of the number of same phonemes to the number of all phonemes is larger than a preset ratio threshold.

### Claim 13 (depends on 1)

A non-transitory computer-readable storage medium, in which computer-executable instructions are stored, wherein the instructions, when executed by a processor, causes the processor to implement the method of claim 1.
