Condivisione della tecnologia

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

I codici demo forniti da Alibaba sono tutti JavaScript, quindi puoi crearne uno tuo. Come riferimento, ho scritto un modulo dattiloscritto per la dettatura delle frasi Alibaba Nls. Il modulo API combinato di VUE3

startClient: avvia la dettatura. Tieni presente che il passaggio successivo deve essere quello di abilitare il riconoscimento e la trasmissione dei dati il ​​prima possibile, altrimenti verrà chiuso dopo 6 secondi.

startRecognition: avvia la transazione di riconoscimento, passa la richiamata di riconoscimento, puoi stampare i caratteri o visualizzarli sullo schermo

sendSound: invia dati PCM binari (formato 16 MHz 16 bit)

stopRecognition: termina la transazione di riconoscimento

/**
 * 阿里语音，一句话识别模块for ccframe
 *
 * 无心跳设计，非长连接推送，因此在需要使用的时候才进行连接
 *
 * @Jim 2024/07/08
 */
import * as utils from '@/utils/index'
import { nextTick } from 'vue'
// import Global from '@/utils/constants'
 
const NLS_SERVER_URL = 'wss://nls-gateway.aliyuncs.com/ws/v1'
const NLS_MODE = 'SpeechRecognizer' // 一句话识别
const WEBSOCKET_MAX_RETRY = 3
const RECONNECT_INTERVAL = 3000
 
interface INlsConfig {
  url?: string
  appkey: string // 应用的key
  token: string // 从服务器获得，要缓存
}
 
let client: (UniNamespace.SocketTask & { readyState?: WsState }) | undefined
const clientId = utils.uuid(utils.UUIDFormat.StandardCompact)
let taskId: string = ''
let config: INlsConfig
let reconnectAttempts = 0
let taskStarted = false
 
enum WsState {
  CONNECTING,
  OPEN,
  CLOSING,
  CLOSED
}
 
/**
 *
 * @param action
 * @returns 请求json
 */
const buildMsg: (action: string, payload: Record<string, any>) => string = (
  action,
  payload = {}
) => {
  if (taskId.length === 0) {
    taskId = utils.uuid(utils.UUIDFormat.StandardCompact)
  }
  const msg = {
    header: {
      message_id: utils.uuid(utils.UUIDFormat.StandardCompact),
      task_id: taskId,
      namespace: NLS_MODE,
      name: action,
      appkey: config.appkey
    },
    payload,
    context: {
      sdk: {
        name: 'nls-wx-sdk',
        version: '0.0.1',
        language: 'wxjs'
      }
    }
  }
  return JSON.stringify(msg, null, 0)
}
 
/**
 * 开启连接，开启后立即要传，否则会被关闭.
 * @param config
 * @param callback
 */
export const startClient = (
  conf?: INlsConfig,
  startCallback?: () => void,
  recognizedCallback?: (text: string) => void
) => {
  if (client && client.readyState !== WsState.CLOSED) {
    // 关闭原连接
    client.close({})
  }
 
  client = uni.connectSocket({
    url: conf.url ?? NLS_SERVER_URL,
    tcpNoDelay: true,
    header: {
      'X-NLS-Token': conf?.token ?? config.token
    },
    success: (res) => {
      if (!config) config = conf
      console.log(`connected to ${NLS_SERVER_URL} success`)
    },
    fail: (res) => {
      console.log(`connect to ${NLS_SERVER_URL} failed:${res.errMsg}`)
    }
  })
  client.readyState = WsState.CONNECTING
 
  client.onMessage((res) => {
    if (typeof res.data === 'string') {
      const msgObj = JSON.parse(res.data)
      switch (msgObj?.header?.name) {
        case 'RecognitionStarted': {
          console.log('started')
          break
        }
        case 'RecognitionResultChanged': {
          if (recognizedCallback) {
            const text = msgObj?.payload?.result
            if (text) {
              recognizedCallback(text)
            }
          }
          console.log('changed')
          break
        }
        case 'RecognitionCompleted': {
          const text = msgObj?.payload?.result
          if (text) {
            recognizedCallback(text)
          }
          taskStarted = false // 结束识别
          break
        }
        case 'TaskFailed': {
          taskStarted = false // 结束识别
          break
        }
      }
    }
    console.log('recv:' + res.data)
  })
 
  client.onOpen(() => {
    reconnectAttempts = 0
    client.readyState = WsState.OPEN
    if (startCallback) nextTick(startCallback)
  })
 
  client.onError((error) => {
    console.error('WebSocket error:', error)
    if (reconnectAttempts < WEBSOCKET_MAX_RETRY) {
      setTimeout(() => startClient(), RECONNECT_INTERVAL)
    } else {
      console.error('Max reconnect attempts reached')
    }
  })
 
  client.onClose(() => {
    client.readyState = WsState.CLOSED
    console.log('connection closed')
  })
}
 
export const startRecognition = () => {
  if (client && client.readyState === WsState.OPEN)
    client.send({
      data: buildMsg('StartRecognition', {
        format: 'opus',
        sample_rate: 16000,
        enable_intermediate_result: true,
        enable_punctuation_prediction: true,
        enable_inverse_text_normalization: true
      }),
      success: (res) => {
        taskStarted = true
      }
    })
}
 
export const stopRecognition = () => {
  if (client && client.readyState === WsState.OPEN)
    client.send({
      data: buildMsg('StopRecognition', {
        format: 'opus',
        sample_rate: 16000,
        enable_intermediate_result: true,
        enable_punctuation_prediction: true,
        enable_inverse_text_normalization: true
      }),
      complete: () => {
        taskStarted = false // 不管是否成功，都不发送音频了
      }
    })
}
 
export const sendSound = (msgBytes: ArrayBuffer) => {
  if (client && client.readyState === WsState.OPEN && taskStarted)
    client.send({
      data: msgBytes,
      success: (res) => {
        console.log('send ' + msgBytes.byteLength + ' success')
      }
    })
}

Per lo strumento uuid di util, vedere il mio articolo precedenteItaliano: Italiano: Italiano: https://mp.csdn.net/mp_blog/creation/editor/140267684Italiano: Italiano: Italiano: https://mp.csdn.net/mp_blog/creation/editor/140267684