技术共享

【UNI-APP】阿里NLS一句话听写typescript模块

2024-07-12

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

阿里提供的demo代码都是javascript,自己捏个轮子。参考着自己写了一个阿里巴巴一句话听写Nls的typescript模块。VUE3的组合式API形式

startClient:开始听写,注意下一步要尽快开启识别和传数据,否则6秒后会关闭

startRecognition:开始识别事务,传入识别回调,可以打印字符或显示到屏幕

sendSound:发送二进制PCM数据(格式16MHz16bit)

stopRecognition:结束识别事务

  1. /**
  2. * 阿里语音,一句话识别模块for ccframe
  3. *
  4. * 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
  5. *
  6. * @Jim 2024/07/08
  7. */
  8. import * as utils from '@/utils/index'
  9. import { nextTick } from 'vue'
  10. // import Global from '@/utils/constants'
  11. const NLS_SERVER_URL = 'wss://nls-gateway.aliyuncs.com/ws/v1'
  12. const NLS_MODE = 'SpeechRecognizer' // 一句话识别
  13. const WEBSOCKET_MAX_RETRY = 3
  14. const RECONNECT_INTERVAL = 3000
  15. interface INlsConfig {
  16. url?: string
  17. appkey: string // 应用的key
  18. token: string // 从服务器获得,要缓存
  19. }
  20. let client: (UniNamespace.SocketTask & { readyState?: WsState }) | undefined
  21. const clientId = utils.uuid(utils.UUIDFormat.StandardCompact)
  22. let taskId: string = ''
  23. let config: INlsConfig
  24. let reconnectAttempts = 0
  25. let taskStarted = false
  26. enum WsState {
  27. CONNECTING,
  28. OPEN,
  29. CLOSING,
  30. CLOSED
  31. }
  32. /**
  33. *
  34. * @param action
  35. * @returns 请求json
  36. */
  37. const buildMsg: (action: string, payload: Record<string, any>) => string = (
  38. action,
  39. payload = {}
  40. ) => {
  41. if (taskId.length === 0) {
  42. taskId = utils.uuid(utils.UUIDFormat.StandardCompact)
  43. }
  44. const msg = {
  45. header: {
  46. message_id: utils.uuid(utils.UUIDFormat.StandardCompact),
  47. task_id: taskId,
  48. namespace: NLS_MODE,
  49. name: action,
  50. appkey: config.appkey
  51. },
  52. payload,
  53. context: {
  54. sdk: {
  55. name: 'nls-wx-sdk',
  56. version: '0.0.1',
  57. language: 'wxjs'
  58. }
  59. }
  60. }
  61. return JSON.stringify(msg, null, 0)
  62. }
  63. /**
  64. * 开启连接,开启后立即要传,否则会被关闭.
  65. * @param config
  66. * @param callback
  67. */
  68. export const startClient = (
  69. conf?: INlsConfig,
  70. startCallback?: () => void,
  71. recognizedCallback?: (text: string) => void
  72. ) => {
  73. if (client && client.readyState !== WsState.CLOSED) {
  74. // 关闭原连接
  75. client.close({})
  76. }
  77. client = uni.connectSocket({
  78. url: conf.url ?? NLS_SERVER_URL,
  79. tcpNoDelay: true,
  80. header: {
  81. 'X-NLS-Token': conf?.token ?? config.token
  82. },
  83. success: (res) => {
  84. if (!config) config = conf
  85. console.log(`connected to ${NLS_SERVER_URL} success`)
  86. },
  87. fail: (res) => {
  88. console.log(`connect to ${NLS_SERVER_URL} failed:${res.errMsg}`)
  89. }
  90. })
  91. client.readyState = WsState.CONNECTING
  92. client.onMessage((res) => {
  93. if (typeof res.data === 'string') {
  94. const msgObj = JSON.parse(res.data)
  95. switch (msgObj?.header?.name) {
  96. case 'RecognitionStarted': {
  97. console.log('started')
  98. break
  99. }
  100. case 'RecognitionResultChanged': {
  101. if (recognizedCallback) {
  102. const text = msgObj?.payload?.result
  103. if (text) {
  104. recognizedCallback(text)
  105. }
  106. }
  107. console.log('changed')
  108. break
  109. }
  110. case 'RecognitionCompleted': {
  111. const text = msgObj?.payload?.result
  112. if (text) {
  113. recognizedCallback(text)
  114. }
  115. taskStarted = false // 结束识别
  116. break
  117. }
  118. case 'TaskFailed': {
  119. taskStarted = false // 结束识别
  120. break
  121. }
  122. }
  123. }
  124. console.log('recv:' + res.data)
  125. })
  126. client.onOpen(() => {
  127. reconnectAttempts = 0
  128. client.readyState = WsState.OPEN
  129. if (startCallback) nextTick(startCallback)
  130. })
  131. client.onError((error) => {
  132. console.error('WebSocket error:', error)
  133. if (reconnectAttempts < WEBSOCKET_MAX_RETRY) {
  134. setTimeout(() => startClient(), RECONNECT_INTERVAL)
  135. } else {
  136. console.error('Max reconnect attempts reached')
  137. }
  138. })
  139. client.onClose(() => {
  140. client.readyState = WsState.CLOSED
  141. console.log('connection closed')
  142. })
  143. }
  144. export const startRecognition = () => {
  145. if (client && client.readyState === WsState.OPEN)
  146. client.send({
  147. data: buildMsg('StartRecognition', {
  148. format: 'opus',
  149. sample_rate: 16000,
  150. enable_intermediate_result: true,
  151. enable_punctuation_prediction: true,
  152. enable_inverse_text_normalization: true
  153. }),
  154. success: (res) => {
  155. taskStarted = true
  156. }
  157. })
  158. }
  159. export const stopRecognition = () => {
  160. if (client && client.readyState === WsState.OPEN)
  161. client.send({
  162. data: buildMsg('StopRecognition', {
  163. format: 'opus',
  164. sample_rate: 16000,
  165. enable_intermediate_result: true,
  166. enable_punctuation_prediction: true,
  167. enable_inverse_text_normalization: true
  168. }),
  169. complete: () => {
  170. taskStarted = false // 不管是否成功,都不发送音频了
  171. }
  172. })
  173. }
  174. export const sendSound = (msgBytes: ArrayBuffer) => {
  175. if (client && client.readyState === WsState.OPEN && taskStarted)
  176. client.send({
  177. data: msgBytes,
  178. success: (res) => {
  179. console.log('send ' + msgBytes.byteLength + ' success')
  180. }
  181. })
  182. }

util的uuid工具见我前一篇文章https://mp.csdn.net/mp_blog/creation/editor/140267684icon-default.png?t=N7T8https://mp.csdn.net/mp_blog/creation/editor/140267684