react-native-baidu-asr

react-native-baidu-asr is a Baidu speech library under React Native, which can perform voice recognition and voice wake-up.

English | 简体中文

Preview

Support

• React Native >= 0.47.0
• Android

Currently, the iOS platform is not implemented. I will fill it up when I have time, as well as voice synthesis.

Install

• RN >= 0.60

1. yarn add react-native-baidu-asr

• RN < 0.60

1. yarn add react-native-baidu-asr

2. react-native link react-native-baidu-asr

Usage

• See for details example

The first is that you have to go to the Baidu Voice Console to create an application, get authentication information: AppID, API Key, Secret Key.

import {
  BaiduAsr,
  StatusCode,
  IBaseData,
  RecognizerResultError,
  RecognizerResultData,
  VolumeData
} from 'react-native-baidu-asr';

// Initialize Baidu speech engine
BaiduAsr.init({
  APP_ID: 'Your authentication information AppID',
  APP_KEY: 'Your authentication information API Key',
  SECRET: 'Your authentication information Secret Key',
});

// Processing recognition results
this.resultListener = BaiduAsr.addResultListener(this.onRecognizerResult);
// Handling wrong results
this.errorListener = BaiduAsr.addErrorListener(this.onRecognizerError);
// Processing volume
this.volumeListener = BaiduAsr.addAsrVolumeListener(this.onAsrVolume);

// Start speech recognition
// For more input parameters, please refer to Baidu Voice Document
// https://ai.baidu.com/ai-doc/SPEECH/bkh07sd0m#asr_start-%E8%BE%93%E5%85%A5%E4%BA%8B%E4%BB%B6%E5%8F%82%E6%95%B0
BaiduAsr.start({
  // Long speech
  VAD_ENDPOINT_TIMEOUT: 0,
  BDS_ASR_ENABLE_LONG_SPEECH: true,
  // Disable punctuation
  DISABLE_PUNCTUATION: true,
});

API

Speech Recognition

Methods

• BaiduAsr.init(options: InitOptions)

Initialize Baidu speech engine

• BaiduAsr.start(options: AsrOptions)

Start speech recognition

• BaiduAsr.stop()

Pause the recording, the SDK will no longer recognize the stopped recording.

• BaiduAsr.cancel()

Cancel the recording, the SDK will cancel this recognition and return to the original state.

• BaiduAsr.release()

Release the resource. If you need to use it again next time, you must call the init method to initialize the engine.

Events

The recognition result callback data has a unified format, similar to the api interface return, with code, msg, and data.

IBaseData The data types are as follows：

interface IBaseData<T = any> {
  /**
   * status code
   */
  code: StatusCode,
  /**
   * message
   */
  msg: string,
  /**
   * data
   */
  data: T
}

• addResultListener(callback: (data: IBaseData<RecognizerResultData | undefined>) => void): EmitterSubscription
  Voice recognition result callback, the event will be triggered continuously during voice recognition，data is of type IBaseData<RecognizerResultData | undefined>，Its value：

  • code：status code
  • msg：message
  • data：Identification data

The data types of data are as follows:

interface RecognizerResultData {
  best_result: string,
  // If there is no accident, the first value is the recognition result
  results_recognition: Array<string>,
  result_type: ResultType,
  origin_result: {
    corpus_no: number,
    err_no: number,
    raf: number,
    result: {
      word: Array<string>
    },
    sn: string
  },
  error: number,
  desc: string
}

• addErrorListener(callback: (data: IBaseData<RecognizerResultError>) => void): EmitterSubscription
  There is an error in speech recognition. The error message is consistent with the Baidu speech document. Its value:

  • code：status code
  • msg：message
  • data：Wrong data

The data types of data are as follows:

interface RecognizerResultError {
  errorCode: number // Error code comparison Baidu voice document https://ai.baidu.com/ai-doc/SPEECH/qk38lxh1q
  subErrorCode: number
  descMessage: string
}

• addAsrVolumeListener(listener: (volume: VolumeData) => void): EmitterSubscription
  The volume of speech recognition. This event will be triggered when the recognized speech changes the volume. volume is of type VolumeData, and its value is:

  • volumePercent: Current volume percentage
  • volume: Current volume

Voice wake

The first is to export wake word , Pre-defined wake words and custom wake words, both need to be exported and used by the wake word evaluation tool.

Methods

• BaiduWakeUp.init(options: InitOptions)

Initialize Baidu speech engine

• BaiduWakeUp.start(options: WakeUpOptions)

Start voice wake up

• BaiduWakeUp.stop()

End voice wakeup.

• BaiduWakeUp.release()

Release the resource. If you need to use it again next time, you must call the init method to initialize the engine.

Events

The wake-up result callback data has a unified format, similar to the api interface return, with code, msg, and data.

The data types of IBaseData are as follows:

interface IBaseData<T = any> {
  /**
   * status code
   */
  code: StatusCode,
  /**
   * message
   */
  msg: string,
  /**
   * data
   */
  data: T
}

• addResultListener(callback: (data: IBaseData<string | undefined>) => void): EmitterSubscription
  Voice wake up result callback, data is IBaseData<string | undefined> type, its value:

  • code：status code
  • msg：message
  • data：Wake word

• addErrorListener(callback: (data: IBaseData<WakeUpResultError>) => void): EmitterSubscription
  There is an error in voice wake-up. The error message is consistent with the Baidu voice document. Its value:

  • code：status code
  • msg：message
  • data：Wrong data

The data types of data are as follows:

interface WakeUpResultError {
  // Error code You can look up the error code against Baidu voice documents https://ai.baidu.com/ai-doc/SPEECH/qk38lxh1q#%E5%94%A4%E9%86%92%E9%94%99%E8%AF%AF%E7%A0%81
  errorCode: number,
  // wrong information
  errorMessage: string,
  // Original error data returned by Baidu Voice
  result: string
}

Contribute

Looking forward to making relevant suggestions, contributions are welcome, thank you star.

Github

License

MIT License