# US 2021/0082431 A1 · Systems and methods for voice recognition

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

- Number: US 2021/0082431 A1
- Document ID: US20210082431A1
- Status: Published application
- Kind: A1
- Application number: 17/103,903
- Filed: 2020-11-24
- Published: 2021-03-18
- Earliest claimed: 2018-05-25
- Assignees: Beijing Didi Infinity Technology and Development Co., Ltd.
- Applicants: Beijing Didi Infinity Technology and Development Co., Ltd.
- Inventors: Rong ZHOU
- Web page: https://patentgrep.com/patent/US20210082431A1

## CPC

- G10L 15/08: Speech classification or search
- G10L 15/22: Procedures used during a speech recognition process, e.g. man-machine dialogue
- G06N 3/04: Architecture, e.g. interconnection topology
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G10L 15/02: Feature extraction for speech recognition; Selection of recognition unit
- G10L 15/16: using artificial neural networks
- G10L 2015/088: {Word spotting}
- G10L 2015/223: {Execution procedure of a spoken command}
- 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

The present disclosure is related to systems and methods for providing voice recognition. The method includes receiving a voice signal including a plurality of frames of voice data. The method also includes determining a voice feature for each frame, the voice feature being related to one or more labels. The method further includes determining one or more scores with respect to the one or more labels based on the voice feature. The method further includes sampling a plurality of frames in a pre-set interval. The method further includes obtaining a score of a label associated with each sampled frame. The method still further includes generating a command to wake up a device based on the obtained scores of the labels associated with the sampled frames.

## Claims

### Claim 1

A system for providing voice recognition, comprising:
at least one storage medium storing a set of instructions; and
at least one processor configured to communicate with the at least one storage medium, wherein when executing the set of instructions, the at least one processor is directed to:
receive a voice signal including a plurality of frames of voice data;
determine a voice feature for each of the plurality of frames, the voice feature being related to one or more labels;
determine one or more scores with respect to the one or more labels based on the voice feature;
sample a plurality of frames in a pre-set interval, the sampled frames corresponding to at least a part of the one or more labels according to a sequence of the one or more labels;
obtain a score of a label associated with each sampled frame; and
generate a command to wake up a device based on the obtained scores of the labels associated with the sampled frames.

### Claim 2 (depends on 1)

The system of claim 1, wherein the at least one processor is further directed to:
perform a smoothing operation on the one or more scores of the one or more labels for each of the plurality of frames.

### Claim 3 (depends on 2)

The system of claim 2, wherein to perform a smoothing operation on one or more scores of one or more labels for each of the plurality of frames, the at least one processor is directed to:
determine a smoothing window with respect to a current frame;
determine at least one frame in the smoothing window associated with the current frame;
determine scores of the one or more labels for the at least one frame;
determine an average score of each of the one or more labels for the current frame based on the scores of the one or more labels for the at least one frame; and
designate the average score of each of the one or more labels for the current frame as the score of each of the one or more labels for the current frame.

### Claim 4 (depends on 1)

The system of claim 1, wherein the one or more labels relate to a wake-up phrase for waking up the device, and the wake-up phrase includes at least one word.

### Claim 5 (depends on 1)

The system of claim 1, wherein to determine one or more scores with respect to the one or more labels based on the one or more voice features, the at least one processor is directed to:
determine a neural network model;
input the one or more voice features corresponding to the plurality of frames into the neural network model; and
generate one or more scores with respect to the one or more labels for each of the one or more voice features.

### Claim 6 (depends on 1)

The system of claim 1, wherein to sample the plurality of frames in a pre-set interval, the at least one processor is directed to:
determine a searching window of a pre-determined width, the pre-determined width of the searching window relating to a number of words in a wake-up phrase; and
determine a number of frames in the searching window, the number of frames corresponding to a first number of labels according to the sequence.

### Claim 7 (depends on 6)

The system of claim 6, wherein to generate a command to wake up a device based on the obtained scores of the labels associated with the sampled frames, the at least one processor is directed to:
determine a final score based on the scores of the one or more labels corresponding to the sampled frames;
determine whether the final score is greater than a threshold; and
in response to the determination that the final score is greater than the threshold,
generate the command to wake up the device.

### Claim 8 (depends on 7)

The system of claim 7, wherein the final score is a radication of a multiplication of the scores of the labels associated with the sampled frames.

### Claim 9 (depends on 7)

The system of claim 7, wherein the at least one processor is further directed to:
in response to the determination that the final score is not greater than the threshold,
move the searching window a step forward.

### Claim 10 (depends on 1)

The system of claim 1, wherein to determine one or more voice features for each of the plurality of frames, the at least one processor is directed to:
transform the voice signal from a time domain to a frequency domain; and
discretize the transformed voice signal to obtain the one or more voice features corresponding to the plurality of frames.

### Claim 11

A method for providing voice recognition implemented on a computing device having one or more processors and one or more storage devices, the method comprising:
receiving a voice signal including a plurality of frames of voice data;
determining a voice feature for each of the plurality of frames, the voice feature being related to one or more labels;
determining one or more scores with respect to the one or more labels based on the voice feature;
sampling a plurality of frames in a pre-set interval, the sampled frames corresponding to at least a part of the one or more labels according to a sequence of the one or more labels;
obtaining a score of a label associated with each sampled frame; and
generating a command to wake up a device based on the obtained scores of the labels associated with the sampled frames.

### Claim 12 (depends on 11)

The method of claim 11, further comprising performing a smoothing operation on the one or more scores of the one or more labels for each of the plurality of frames.

### Claim 13 (depends on 12)

The method of claim 12, wherein performing a smoothing operation on one or more scores of one or more labels for each of the plurality of frames comprises:
determining a smoothing window with respect to a current frame;
determining at least one frame in the smoothing window associated with the current frame;
determining scores of the one or more labels for the at least one frame;
determining an average score of each of the one or more labels for the current frame based on the scores of the one or more labels for the at least one frame; and
designating the average score of each of the one or more labels for the current frame as the score of each of the one or more labels for the current frame.

### Claim 14 (depends on 11)

The method of claim 11, wherein the one or more labels relate to a wake-up phrase for waking up the device, and the wake-up phrase includes at least one word.

### Claim 15 (depends on 11)

The method of claim 11, wherein determining one or more scores with respect to the one or more labels based on the one or more voice features comprises:
determining a neural network model;
inputting the one or more voice features corresponding to the plurality of frames into the neural network model; and
generating one or more scores with respect to the one or more labels for each of the one or more voice features.

### Claim 16 (depends on 11)

The method of claim 11, wherein sampling the plurality of frames in a pre-set interval comprises:
determining a searching window of a pre-determined width, the pre-determined width of the searching window relating to a number of words in a wake-up phrase; and
determining a number of frames in the searching window, the number of frames corresponding to a first number of labels according to the sequence.

### Claim 17 (depends on 16)

The method of claim 16, wherein generating a command to wake up a device based on the obtained scores of the labels associated with the sampled frames comprises:
determining a final score based on the scores of the one or more labels corresponding to the sampled frames;
determining whether the final score is greater than a threshold; and
in response to the determination that the final score is greater than the threshold,
generating the command to wake up the device.

### Claim 18 (depends on 17)

The method of claim 17, further comprising:
in response to the determination that the final score is not greater than the threshold,
moving the searching window a step forward.

### Claim 19 (depends on 11)

The method of claim 11, wherein determining one or more voice features for each of the plurality of frames comprises:
transforming the voice signal from a time domain to a frequency domain; and
discretizing the transformed voice signal to obtain the one or more voice features corresponding to the plurality of frames.

### Claim 20

A non-transitory computer readable medium, comprising at least one set of instructions for providing voice recognition, wherein when executed by one or more processors of a computing device, the at least one set of instructions causes the computing device to perform a method, the method comprising:
receiving a voice signal including a plurality of frames of voice data;
determining a voice feature for each of the plurality of frames, the voice feature being related to one or more labels;
determining one or more scores with respect to the one or more labels based on the voice feature;
sampling a plurality of frames in a pre-set interval, the sampled frames corresponding to at least a part of the one or more labels according to a sequence of the one or more labels;
obtaining a score of a label associated with each sampled frame; and
generating a command to wake up a device based on the obtained scores of the labels associated with the sampled frames.
