HarmonyOS NEXT AI 基础语音服务 - 语音输入
作者：zhousg
- 2025-03-30 北京
本文字数：2843 字
阅读完需：约 9 分钟
案例描述
这是一个基于 AI 基础语音服务实现的实时语音转文字案例，通过麦克风采集音频并实时转换为文本。
实现步骤：
1. 导入必要模块
import { speechRecognizer } from '@kit.CoreSpeechKit'import { abilityAccessCtrl } from '@kit.AbilityKit'import { promptAction } from '@kit.ArkUI'
复制代码
2. 申请麦克风权限
async requestPermissions() { const atManager = abilityAccessCtrl.createAtManager(); const res = await atManager.requestPermissionsFromUser(getContext(), ['ohos.permission.MICROPHONE']) this.hasPermissions = res.authResults.every(grantStatus => grantStatus === abilityAccessCtrl.GrantStatus.PERMISSION_GRANTED)}
复制代码
3. 初始化语音识别引擎
async startRecord() { if (canIUse('SystemCapability.AI.SpeechRecognizer')) { this.asrEngine = await speechRecognizer.createEngine({ language: 'zh-CN', online: 1 }) // ...引擎初始化代码... }}
复制代码
"requestPermissions": [ { "name": "ohos.permission.MICROPHONE", "reason": "$string:EntryAbility_label", "usedScene": {} } ],
复制代码
4. 设置语音识别回调
this.asrEngine.setListener({ onResult(sessionId: string, result: speechRecognizer.SpeechRecognitionResult) { _this.text = result.result // 实时更新识别结果 if (result.isLast) { // 识别结束处理 _this.isRecording = false } }, // ...其他回调方法...})
复制代码
5. 配置音频参数
const audioParam: speechRecognizer.AudioInfo = { audioType: 'pcm', // 音频格式 sampleRate: 16000, // 采样率 soundChannel: 1, // 单声道 sampleBit: 16 // 采样位数}
复制代码
落地代码：
完整组件结构
@Entry@ComponentV2struct SpeechRecognizer { @Local isRecording: boolean = false @Local text: string = '' hasPermissions: boolean = false asrEngine?: speechRecognizer.SpeechRecognitionEngine
// 组件生命周期方法 aboutToAppear(): void { this.requestPermissions() }
// ...其他方法实现...}
复制代码
UI 交互设计
build() { Column() { // 文本展示区域 Row() { Text(this.text) .width('100%') }
// 长按语音按钮 Button(this.isRecording ? '开始 说话' : '按住 说话') .gesture(LongPressGesture() .onAction(() => this.startRecord()) .onActionEnd(() => this.closeRecord())) }}
复制代码
总结梳理：
核心点
权限管理：使用
AbilityKit动态申请麦克风权限引擎生命周期：在组件
aboutToAppear初始化，操作结束及时释放资源语音识别流程：
创建识别引擎
配置音频参数（PCM 格式/16K 采样率）
设置结果回调实时更新 UI
长按手势控制识别启停
异常处理：通过
promptAction提示权限异常和设备忙状态
完整代码
import { speechRecognizer } from '@kit.CoreSpeechKit'import { abilityAccessCtrl } from '@kit.AbilityKit'import { promptAction } from '@kit.ArkUI'
@Entry@ComponentV2struct SpeechRecognizer { @Local isRecording: boolean = false @Local text: string = '' hasPermissions: boolean = false asrEngine?: speechRecognizer.SpeechRecognitionEngine
aboutToAppear(): void { // 获取麦克风权限 this.requestPermissions() }
async requestPermissions() { const atManager = abilityAccessCtrl.createAtManager(); const res = await atManager.requestPermissionsFromUser(getContext(), ['ohos.permission.MICROPHONE']) this.hasPermissions = res.authResults.every(grantStatus => grantStatus === abilityAccessCtrl.GrantStatus.PERMISSION_GRANTED) }
// 开始麦克风识别 async startRecord() { if (canIUse('SystemCapability.AI.SpeechRecognizer')) { if (!this.hasPermissions) { return promptAction.showToast({ message: '麦克风未授权' }) } if (this.isRecording) { return promptAction.showToast({ message: '正在录制...' }) } this.isRecording = true this.asrEngine = await speechRecognizer.createEngine({ language: 'zh-CN', online: 1 }) const _this = this this.asrEngine.setListener({ onStart(sessionId: string, eventMessage: string) { }, onEvent(sessionId: string, eventCode: number, eventMessage: string) { }, onResult(sessionId: string, result: speechRecognizer.SpeechRecognitionResult) { _this.text = result.result if (result.isLast) { _this.isRecording = false } }, onComplete(sessionId: string, eventMessage: string) { }, onError(sessionId: string, errorCode: number, errorMessage: string) { } }) const audioParam: speechRecognizer.AudioInfo = { audioType: 'pcm', sampleRate: 16000, soundChannel: 1, sampleBit: 16 } const extraParam: Record<string, Object> = { "recognitionMode": 0, "vadBegin": 2000, "vadEnd": 3000, "maxAudioDuration": 20000 } const recognizerParams: speechRecognizer.StartParams = { sessionId: '10000', audioInfo: audioParam, extraParams: extraParam } this.asrEngine.startListening(recognizerParams) } }
async closeRecord() { if (canIUse('SystemCapability.AI.SpeechRecognizer')) { this.asrEngine?.finish('10000') this.asrEngine?.cancel('10000') this.asrEngine?.shutdown() } }
build() { Column() { Row() { Text(this.text) .width('100%') .lineHeight(32) } .alignItems(VerticalAlign.Top) .width('100%') .layoutWeight(1)
Button(this.isRecording ? '开始 说话' : '按住 说话') .width('100%') .gesture(LongPressGesture() .onAction(() => { this.startRecord() }) .onActionEnd(() => { this.closeRecord() }) .onActionCancel(() => { this.closeRecord() }))
} .padding(15) .height('100%') .width('100%') }}
复制代码
划线
评论
复制
发布于: 刚刚阅读数: 3
版权声明: 本文为 InfoQ 作者【zhousg】的原创文章。
原文链接:【http://xie.infoq.cn/article/9d75a587adabe837006dcbdb4】。文章转载请联系作者。
zhousg
关注
还未添加个人签名 2025-03-14 加入
还未添加个人简介
评论