recorder-client 1.0.6 → 1.1.0

package/README.md

@@ -14,11 +14,11 @@ npm i recorder-client
 
 ### API
 
-|  property  |  type   |                                 description                                 |  default  |
-| :--------: | :-----: | :-------------------------------------------------------------------------: | :-------: |
-| sampleRate | number  |                             audio sampling rate                             | undefined |
-| chunkSize  | number  | number of samples in the audio data block (length of the audio buffer zone) | undefined |
-| ignoreMute | boolean |                                 ignore mute                                 | undefined |
+|  property  |  type  |                                 description                                 |  default  |
+| :--------: | :----: | :-------------------------------------------------------------------------: | :-------: |
+| sampleRate | number |                             audio sampling rate                             | undefined |
+| chunkSize  | number | number of samples in the audio data block (length of the audio buffer zone) | undefined |
+|    vad     |  Vad   |          determine whether the audio clip contains spoken content           | undefined |
 
 ### Event
 
@@ -36,12 +36,11 @@ npm i recorder-client
 
 ### Helper
 
-| property |                       type                       |              description              |
-| :------: | :----------------------------------------------: | :-----------------------------------: |
-| pcmMerge |        (pcms: Int16Array[]) => Int16Array        |      merge multiple pcm streams       |
-| pcmToWav |  (pcm: Int16Array, sampleRate: number) => Blob   |          convert PCM to WAV           |
-| parseWav |         (buffer: ArrayBuffer) => object          | obtain information about the wav file |
-| isSpeak  | (pcm: Int16Array, sampleRate: number) => boolean | check if there is any input of audio  |
+| property |                     type                      |              description              |
+| :------: | :-------------------------------------------: | :-----------------------------------: |
+| pcmMerge |      (pcms: Int16Array[]) => Int16Array       |      merge multiple pcm streams       |
+| pcmToWav | (pcm: Int16Array, sampleRate: number) => Blob |          convert PCM to WAV           |
+| parseWav |        (buffer: ArrayBuffer) => object        | obtain information about the wav file |
 
 ### Example
 
@@ -69,6 +68,37 @@ const onClick = () => {
 };
 ```
 
+> VAD
+
+```ts
+import { Recorder } from 'recorder-client';
+
+type VadScene =
+  | 'extreme-noise'
+  | 'high-noise'
+  | 'live-stream'
+  | 'normal-indoor'
+  | 'quiet-studio'
+  | 'voice-assistant';
+
+const recorder = new Recorder({
+  sampleRate: 16000,
+  chunkSize: 1900,
+  vad: 'normal-indoor',
+});
+// custom
+const recorder = new Recorder({
+  sampleRate: 16000,
+  chunkSize: 1900,
+  vad: {
+    noiseCoefficient: 2.3,
+    voiceRatioThreshold: 0.5,
+    noiseFrameCount: 5,
+    frameDurationMs: 20,
+  },
+});
+```
+
 > pcmMerge
 
 ```ts
@@ -123,18 +153,3 @@ const getWav = async (file: File) => {
   const wav: Wav = Recorder.parseWav(buffer);
 };
 ```
-
-> isSpeak
-
-```ts
-import { Recorder } from 'recorder-client';
-
-const recorder = new Recorder({
-  sampleRate: 16000,
-  chunkSize: 1900,
-});
-
-recorder.ondataavailable = (pcm) => {
-  const flag = Recorder.isSpeak(pcm, 16000);
-};
-```

package/dist/index.cjs.js

@@ -1,8 +1,8 @@
 'use strict';
 
-//参考 AudioWorkletProcessor 单声道 音频采样率是16000 使用pcm 1.9kb发送一次数据
+//AudioWorkletProcessor 单声道 音频采样率是16000 使用pcm 1.9kb发送一次数据
 function audioProcessorJs() {
-  const code = `
+    const code = `
 class AudioProcessor extends AudioWorkletProcessor {
   constructor({ processorOptions }) {
     super();
@@ -56,251 +56,323 @@ class AudioProcessor extends AudioWorkletProcessor {
 
 registerProcessor('audio-processor', AudioProcessor);
 `;
-  const blob = new Blob([code], { type: 'application/javascript' });
-  return URL.createObjectURL(blob);
+    const blob = new Blob([code], { type: 'application/javascript' });
+    return URL.createObjectURL(blob);
 }
+/** 场景化预设配置表 */
+const SCENE_PRESETS = {
+    'quiet-studio': {
+        noiseCoefficient: 1.6,
+        voiceRatioThreshold: 0.4,
+        noiseFrameCount: 3,
+        frameDurationMs: 20,
+    },
+    'normal-indoor': {
+        noiseCoefficient: 2.3,
+        voiceRatioThreshold: 0.5,
+        noiseFrameCount: 5,
+        frameDurationMs: 20,
+    },
+    'high-noise': {
+        noiseCoefficient: 3.8,
+        voiceRatioThreshold: 0.6,
+        noiseFrameCount: 8,
+        frameDurationMs: 15, // 更短的帧提升实时性
+    },
+    'extreme-noise': {
+        noiseCoefficient: 4.5,
+        voiceRatioThreshold: 0.7,
+        noiseFrameCount: 10,
+        frameDurationMs: 10, // 极短帧快速响应噪声变化
+    },
+    'voice-assistant': {
+        noiseCoefficient: 2.0,
+        voiceRatioThreshold: 0.3, // 低占比触发，提升唤醒灵敏度
+        noiseFrameCount: 2,
+        frameDurationMs: 20,
+    },
+    'live-stream': {
+        noiseCoefficient: 2.8,
+        voiceRatioThreshold: 0.45,
+        noiseFrameCount: 6,
+        frameDurationMs: 20,
+    },
+};
 class Recorder {
-  constructor(options) {
-    Object.defineProperty(this, '_stream', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, '_audioContext', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, '_source', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, '_workletNode', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, '_pcmData', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, '_options', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, 'ondataavailable', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, 'onpause', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, 'onresume', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, 'onstart', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, 'onstop', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    this._options = options;
-    this._pcmData = [];
-  }
-  async start() {
-    const { sampleRate, chunkSize, ignoreMute } = this._options;
-    this._stream = await navigator.mediaDevices.getUserMedia({
-      audio: true,
-    });
-    this._audioContext = new AudioContext({ sampleRate });
-    this._source = this._audioContext.createMediaStreamSource(this._stream);
-    await this._audioContext.audioWorklet.addModule(audioProcessorJs(), { credentials: 'include' });
-    this._workletNode = new AudioWorkletNode(this._audioContext, 'audio-processor', {
-      channelCount: 1, //单声道
-      channelCountMode: 'explicit', //输出通道数由 channelCount 显式指定
-      channelInterpretation: 'discrete', //通道是独立的音频流，没有特定方向（适合单声道处理）
-      processorOptions: { chunkSize },
-    });
-    this._source.connect(this._workletNode);
-    this._workletNode.connect(this._audioContext.destination);
-    this._workletNode.port.onmessage = (event) => {
-      const { e, data } = event.data;
-      if (e === 'data') {
-        if (ignoreMute) {
-          if (Recorder.isSpeak(data, sampleRate)) {
-            this._pcmData.push(data);
-            this.ondataavailable?.(data);
-          }
-        } else {
-          this._pcmData.push(data);
-          this.ondataavailable?.(data);
-        }
-      }
-    };
-    this.onstart?.();
-  }
-  pause() {
-    this._workletNode?.port.postMessage('pause');
-    this.onpause?.();
-  }
-  resume() {
-    this._workletNode?.port.postMessage('resume');
-    this.onresume?.();
-  }
-  stop() {
-    this._workletNode?.port.postMessage('stop');
-    this._stream?.getTracks().forEach((track) => {
-      track.stop();
-    });
-    this._source?.disconnect();
-    this._workletNode?.disconnect();
-    this._stream = undefined;
-    this._source = undefined;
-    this._workletNode = undefined;
-    this._audioContext = undefined;
-    this.onstop?.(Recorder.pcmMerge(this._pcmData));
-    this._pcmData = [];
-  }
-  static pcmMerge(arrays) {
-    // 计算总长度
-    const totalLength = arrays.reduce((sum, arr) => sum + arr.length, 0);
-    // 创建合并后的 Int16Array
-    const merged = new Int16Array(totalLength);
-    let offset = 0;
-    for (const arr of arrays) {
-      // 将当前数组复制到目标数组的 offset 位置
-      merged.set(arr, offset);
-      offset += arr.length;
+    constructor(options) {
+        Object.defineProperty(this, "_stream", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "_audioContext", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "_source", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "_workletNode", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "_pcmData", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "_options", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "ondataavailable", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "onpause", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "onresume", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "onstart", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "onstop", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        this._options = options;
+        this._pcmData = [];
     }
-    return merged;
-  }
-  static pcmToWav(data, sampleRate) {
-    function writeString(view, offset, type) {
-      for (let i = 0; i < type.length; i++) {
-        view.setUint8(offset + i, type.charCodeAt(i));
-      }
+    async start() {
+        const { sampleRate, chunkSize, vad } = this._options;
+        this._stream = await navigator.mediaDevices.getUserMedia({
+            audio: true,
+        });
+        this._audioContext = new AudioContext({ sampleRate });
+        this._source = this._audioContext.createMediaStreamSource(this._stream);
+        await this._audioContext.audioWorklet.addModule(audioProcessorJs(), { credentials: 'include' });
+        this._workletNode = new AudioWorkletNode(this._audioContext, 'audio-processor', {
+            channelCount: 1, //单声道
+            channelCountMode: 'explicit', //输出通道数由 channelCount 显式指定
+            channelInterpretation: 'discrete', //通道是独立的音频流，没有特定方向（适合单声道处理）
+            processorOptions: { chunkSize },
+        });
+        this._source.connect(this._workletNode);
+        this._workletNode.connect(this._audioContext.destination);
+        this._workletNode.port.onmessage = (event) => {
+            const { e, data } = event.data;
+            if (e === 'data') {
+                if (vad) {
+                    if (Recorder.isSpeak(data, sampleRate, vad)) {
+                        this._pcmData.push(data);
+                        this.ondataavailable?.(data);
+                    }
+                }
+                else {
+                    this._pcmData.push(data);
+                    this.ondataavailable?.(data);
+                }
+            }
+        };
+        this.onstart?.();
     }
-    const numChannels = 1; // 单声道
-    const bytesPerSample = 2; // 16 位
-    const dataLength = data.length;
-    // 计算总文件大小
-    const fileSize = 44 + dataLength * bytesPerSample;
-    // 创建 ArrayBuffer
-    const buffer = new ArrayBuffer(fileSize);
-    const view = new DataView(buffer);
-    // 写入 RIFF 头
-    writeString(view, 0, 'RIFF');
-    view.setUint32(4, fileSize - 8, true); // 文件总长度 - 8
-    writeString(view, 8, 'WAVE');
-    // 写入 fmt 子块
-    writeString(view, 12, 'fmt ');
-    view.setUint32(16, 16, true); // 子块大小
-    view.setUint16(20, 1, true); // 音频格式 (PCM)
-    view.setUint16(22, numChannels, true); // 通道数
-    view.setUint32(24, sampleRate, true); // 采样率
-    view.setUint32(28, sampleRate * numChannels * bytesPerSample, true); // 字节率
-    view.setUint16(32, numChannels * bytesPerSample, true); // 块对齐
-    view.setUint16(34, bytesPerSample * 8, true); // 位深度 (16 位)
-    // 写入 data 子块
-    writeString(view, 36, 'data');
-    view.setUint32(40, dataLength * bytesPerSample, true); // 数据大小
-    // 写入 PCM 数据
-    for (let i = 0; i < dataLength; i++) {
-      view.setInt16(44 + i * bytesPerSample, data[i], true); // 小端序
+    pause() {
+        this._workletNode?.port.postMessage('pause');
+        this.onpause?.();
     }
-    const wavBlob = new Blob([buffer], { type: 'audio/wav' });
-    return wavBlob;
-  }
-  static isSpeak(int16Array, sampleRate) {
-    const frameSize = Math.floor(0.02 * sampleRate); // 20ms 帧长度
-    const threshold = 0.01; // 能量阈值（需根据实际调整）
-    let speakingFrames = 0;
-    const frameCount = Math.ceil(int16Array.length / frameSize);
-    for (let i = 0; i < frameCount; i++) {
-      const start = i * frameSize;
-      const end = Math.min(start + frameSize, int16Array.length);
-      const frame = int16Array.slice(start, end);
-      let sum = 0;
-      for (let j = 0; j < frame.length; j++) {
-        const sample = frame[j] / 32768; // 归一化到 [-1, 1]
-        sum += sample * sample;
-      }
-      const rms = Math.sqrt(sum / frame.length);
-      if (rms > threshold) {
-        speakingFrames++;
-      }
+    resume() {
+        this._workletNode?.port.postMessage('resume');
+        this.onresume?.();
     }
-    // 如果超过 50% 的帧有语音，认为在说话
-    return speakingFrames / frameCount > 0.5;
-  }
-  static parseWav(buffer) {
-    // 将32位无符号整数转换为4字节字符串
-    function uint32ToChars(uint32) {
-      return String.fromCharCode(
-        (uint32 >> 24) & 0xff,
-        (uint32 >> 16) & 0xff,
-        (uint32 >> 8) & 0xff,
-        (uint32 >> 0) & 0xff,
-      );
+    stop() {
+        this._workletNode?.port.postMessage('stop');
+        this._stream?.getTracks().forEach((track) => {
+            track.stop();
+        });
+        this._source?.disconnect();
+        this._workletNode?.disconnect();
+        this._stream = undefined;
+        this._source = undefined;
+        this._workletNode = undefined;
+        this._audioContext = undefined;
+        this.onstop?.(Recorder.pcmMerge(this._pcmData));
+        this._pcmData = [];
     }
-    const dataView = new DataView(buffer);
-    let offset = 0;
-    // 1. 验证RIFF头
-    const riff = dataView.getUint32(offset, false); // 大端序
-    if (riff !== 0x52494646) {
-      // 'RIFF' in hex
-      throw new Error('Not a RIFF file');
+    static pcmMerge(arrays) {
+        // 计算总长度
+        const totalLength = arrays.reduce((sum, arr) => sum + arr.length, 0);
+        // 创建合并后的 Int16Array
+        const merged = new Int16Array(totalLength);
+        let offset = 0;
+        for (const arr of arrays) {
+            // 将当前数组复制到目标数组的 offset 位置
+            merged.set(arr, offset);
+            offset += arr.length;
+        }
+        return merged;
     }
-    offset += 4;
-    const fileSize = dataView.getUint32(offset, true); // 小端序
-    offset += 4;
-    const wave = dataView.getUint32(offset, false); // 大端序
-    if (wave !== 0x57415645) {
-      // 'WAVE' in hex
-      throw new Error('Not a WAVE file');
+    static pcmToWav(data, sampleRate) {
+        function writeString(view, offset, type) {
+            for (let i = 0; i < type.length; i++) {
+                view.setUint8(offset + i, type.charCodeAt(i));
+            }
+        }
+        const numChannels = 1; // 单声道
+        const bytesPerSample = 2; // 16 位
+        const dataLength = data.length;
+        // 计算总文件大小
+        const fileSize = 44 + dataLength * bytesPerSample;
+        // 创建 ArrayBuffer
+        const buffer = new ArrayBuffer(fileSize);
+        const view = new DataView(buffer);
+        // 写入 RIFF 头
+        writeString(view, 0, 'RIFF');
+        view.setUint32(4, fileSize - 8, true); // 文件总长度 - 8
+        writeString(view, 8, 'WAVE');
+        // 写入 fmt 子块
+        writeString(view, 12, 'fmt ');
+        view.setUint32(16, 16, true); // 子块大小
+        view.setUint16(20, 1, true); // 音频格式 (PCM)
+        view.setUint16(22, numChannels, true); // 通道数
+        view.setUint32(24, sampleRate, true); // 采样率
+        view.setUint32(28, sampleRate * numChannels * bytesPerSample, true); // 字节率
+        view.setUint16(32, numChannels * bytesPerSample, true); // 块对齐
+        view.setUint16(34, bytesPerSample * 8, true); // 位深度 (16 位)
+        // 写入 data 子块
+        writeString(view, 36, 'data');
+        view.setUint32(40, dataLength * bytesPerSample, true); // 数据大小
+        // 写入 PCM 数据
+        for (let i = 0; i < dataLength; i++) {
+            view.setInt16(44 + i * bytesPerSample, data[i], true); // 小端序
+        }
+        const wavBlob = new Blob([buffer], { type: 'audio/wav' });
+        return wavBlob;
     }
-    offset += 4;
-    // 2. 遍历块，查找'fmt '块
-    while (offset < buffer.byteLength) {
-      const chunkId = dataView.getUint32(offset, false); // 大端序
-      const chunkSize = dataView.getUint32(offset + 4, true); // 小端序
-      const chunkIdStr = uint32ToChars(chunkId);
-      if (chunkIdStr === 'fmt ') {
-        // 3. 解析'fmt '块数据
-        const audioFormat = dataView.getUint16(offset + 8, true); // 小端序
-        const numChannels = dataView.getUint16(offset + 10, true);
-        const sampleRate = dataView.getUint32(offset + 12, true);
-        const byteRate = dataView.getUint32(offset + 16, true);
-        return { numChannels, sampleRate, byteRate, fileSize, audioFormat };
-      }
-      // 移动到下一个块
-      offset += 8 + chunkSize;
+    /**
+     * 语音活动检测（VAD）：判断音频片段是否包含说话内容
+     * @param int16Array 原始PCM音频数据（16位整型）
+     * @param sampleRate 采样率（如16000、44100）
+     * @param vad 场景类型或自定义配置
+     * @returns 是否检测到语音
+     */
+    static isSpeak(int16Array, sampleRate, vad) {
+        // 解析配置：场景预设优先，自定义配置可覆盖
+        let options;
+        if (typeof vad === 'string') {
+            options = { ...SCENE_PRESETS[vad] };
+        }
+        else {
+            // 默认使用普通室内场景配置，再合并自定义选项
+            options = { ...SCENE_PRESETS['normal-indoor'], ...vad };
+        }
+        // 计算帧长度（处理自定义帧时长）
+        const frameSize = Math.floor(((options.frameDurationMs ?? 20) * sampleRate) / 1000);
+        const totalFrameCount = Math.ceil(int16Array.length / frameSize);
+        let speakingFrameCount = 0;
+        let noiseRmsSum = 0;
+        // --------------------------
+        // 1. 计算背景噪声基准RMS
+        // --------------------------
+        const actualNoiseFrameCount = Math.min(options.noiseFrameCount, totalFrameCount);
+        for (let i = 0; i < actualNoiseFrameCount; i++) {
+            const start = i * frameSize;
+            const end = Math.min(start + frameSize, int16Array.length);
+            const frame = int16Array.subarray(start, end);
+            let sum = 0;
+            for (let j = 0; j < frame.length; j++) {
+                const sample = frame[j] / 32768; // 归一化到[-1, 1]
+                sum += sample * sample;
+            }
+            noiseRmsSum += Math.sqrt(sum / frame.length);
+        }
+        const baseNoiseRms = noiseRmsSum / actualNoiseFrameCount;
+        const dynamicThreshold = baseNoiseRms * options.noiseCoefficient;
+        // --------------------------
+        // 2. 逐帧检测语音
+        // --------------------------
+        for (let i = 0; i < totalFrameCount; i++) {
+            const start = i * frameSize;
+            const end = Math.min(start + frameSize, int16Array.length);
+            const frame = int16Array.subarray(start, end);
+            let sum = 0;
+            for (let j = 0; j < frame.length; j++) {
+                const sample = frame[j] / 32768;
+                sum += sample * sample;
+            }
+            const frameRms = Math.sqrt(sum / frame.length);
+            if (frameRms > dynamicThreshold) {
+                speakingFrameCount++;
+            }
+        }
+        // 3. 判断是否超过语音占比阈值
+        return speakingFrameCount / totalFrameCount > options.voiceRatioThreshold;
+    }
+    static parseWav(buffer) {
+        // 将32位无符号整数转换为4字节字符串
+        function uint32ToChars(uint32) {
+            return String.fromCharCode((uint32 >> 24) & 0xff, (uint32 >> 16) & 0xff, (uint32 >> 8) & 0xff, (uint32 >> 0) & 0xff);
+        }
+        const dataView = new DataView(buffer);
+        let offset = 0;
+        // 1. 验证RIFF头
+        const riff = dataView.getUint32(offset, false); // 大端序
+        if (riff !== 0x52494646) {
+            // 'RIFF' in hex
+            throw new Error('Not a RIFF file');
+        }
+        offset += 4;
+        const fileSize = dataView.getUint32(offset, true); // 小端序
+        offset += 4;
+        const wave = dataView.getUint32(offset, false); // 大端序
+        if (wave !== 0x57415645) {
+            // 'WAVE' in hex
+            throw new Error('Not a WAVE file');
+        }
+        offset += 4;
+        // 2. 遍历块，查找'fmt '块
+        while (offset < buffer.byteLength) {
+            const chunkId = dataView.getUint32(offset, false); // 大端序
+            const chunkSize = dataView.getUint32(offset + 4, true); // 小端序
+            const chunkIdStr = uint32ToChars(chunkId);
+            if (chunkIdStr === 'fmt ') {
+                // 3. 解析'fmt '块数据
+                const audioFormat = dataView.getUint16(offset + 8, true); // 小端序
+                const numChannels = dataView.getUint16(offset + 10, true);
+                const sampleRate = dataView.getUint32(offset + 12, true);
+                const byteRate = dataView.getUint32(offset + 16, true);
+                return { numChannels, sampleRate, byteRate, fileSize, audioFormat };
+            }
+            // 移动到下一个块
+            offset += 8 + chunkSize;
+        }
+        throw new Error('fmt chunk not found');
     }
-    throw new Error('fmt chunk not found');
-  }
 }
 
 exports.Recorder = Recorder;

package/dist/index.d.ts

@@ -1,7 +1,17 @@
+/** 预定义的VAD场景类型 */
+type VadScene = 'extreme-noise' | 'high-noise' | 'live-stream' | 'normal-indoor' | 'quiet-studio' | 'voice-assistant';
+/** VAD配置选项类型 */
+interface VadOptions {
+    noiseCoefficient: number;
+    voiceRatioThreshold: number;
+    noiseFrameCount: number;
+    frameDurationMs?: number;
+}
+type Vad = Partial<VadOptions> | VadScene;
 interface RecorderConstructor {
     sampleRate: number;
     chunkSize: number;
-    ignoreMute?: boolean;
+    vad?: Vad;
 }
 declare class Recorder {
     private _stream?;
@@ -22,7 +32,14 @@ declare class Recorder {
     stop(): void;
     static pcmMerge(arrays: Int16Array[]): Int16Array;
     static pcmToWav(data: Int16Array, sampleRate: number): Blob;
-    static isSpeak(int16Array: Int16Array, sampleRate: number): boolean;
+    /**
+     * 语音活动检测（VAD）：判断音频片段是否包含说话内容
+     * @param int16Array 原始PCM音频数据（16位整型）
+     * @param sampleRate 采样率（如16000、44100）
+     * @param vad 场景类型或自定义配置
+     * @returns 是否检测到语音
+     */
+    static isSpeak(int16Array: Int16Array, sampleRate: number, vad?: Vad): boolean;
     static parseWav(buffer: ArrayBuffer): {
         numChannels: number;
         sampleRate: number;
@@ -33,4 +50,4 @@ declare class Recorder {
 }
 
 export { Recorder };
-export type { RecorderConstructor };
+export type { RecorderConstructor, Vad, VadOptions, VadScene };

package/dist/index.esm.js

@@ -1,6 +1,6 @@
-//参考 AudioWorkletProcessor 单声道 音频采样率是16000 使用pcm 1.9kb发送一次数据
+//AudioWorkletProcessor 单声道 音频采样率是16000 使用pcm 1.9kb发送一次数据
 function audioProcessorJs() {
-  const code = `
+    const code = `
 class AudioProcessor extends AudioWorkletProcessor {
   constructor({ processorOptions }) {
     super();
@@ -54,251 +54,323 @@ class AudioProcessor extends AudioWorkletProcessor {
 
 registerProcessor('audio-processor', AudioProcessor);
 `;
-  const blob = new Blob([code], { type: 'application/javascript' });
-  return URL.createObjectURL(blob);
+    const blob = new Blob([code], { type: 'application/javascript' });
+    return URL.createObjectURL(blob);
 }
+/** 场景化预设配置表 */
+const SCENE_PRESETS = {
+    'quiet-studio': {
+        noiseCoefficient: 1.6,
+        voiceRatioThreshold: 0.4,
+        noiseFrameCount: 3,
+        frameDurationMs: 20,
+    },
+    'normal-indoor': {
+        noiseCoefficient: 2.3,
+        voiceRatioThreshold: 0.5,
+        noiseFrameCount: 5,
+        frameDurationMs: 20,
+    },
+    'high-noise': {
+        noiseCoefficient: 3.8,
+        voiceRatioThreshold: 0.6,
+        noiseFrameCount: 8,
+        frameDurationMs: 15, // 更短的帧提升实时性
+    },
+    'extreme-noise': {
+        noiseCoefficient: 4.5,
+        voiceRatioThreshold: 0.7,
+        noiseFrameCount: 10,
+        frameDurationMs: 10, // 极短帧快速响应噪声变化
+    },
+    'voice-assistant': {
+        noiseCoefficient: 2.0,
+        voiceRatioThreshold: 0.3, // 低占比触发，提升唤醒灵敏度
+        noiseFrameCount: 2,
+        frameDurationMs: 20,
+    },
+    'live-stream': {
+        noiseCoefficient: 2.8,
+        voiceRatioThreshold: 0.45,
+        noiseFrameCount: 6,
+        frameDurationMs: 20,
+    },
+};
 class Recorder {
-  constructor(options) {
-    Object.defineProperty(this, '_stream', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, '_audioContext', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, '_source', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, '_workletNode', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, '_pcmData', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, '_options', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, 'ondataavailable', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, 'onpause', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, 'onresume', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, 'onstart', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    Object.defineProperty(this, 'onstop', {
-      enumerable: true,
-      configurable: true,
-      writable: true,
-      value: void 0,
-    });
-    this._options = options;
-    this._pcmData = [];
-  }
-  async start() {
-    const { sampleRate, chunkSize, ignoreMute } = this._options;
-    this._stream = await navigator.mediaDevices.getUserMedia({
-      audio: true,
-    });
-    this._audioContext = new AudioContext({ sampleRate });
-    this._source = this._audioContext.createMediaStreamSource(this._stream);
-    await this._audioContext.audioWorklet.addModule(audioProcessorJs(), { credentials: 'include' });
-    this._workletNode = new AudioWorkletNode(this._audioContext, 'audio-processor', {
-      channelCount: 1, //单声道
-      channelCountMode: 'explicit', //输出通道数由 channelCount 显式指定
-      channelInterpretation: 'discrete', //通道是独立的音频流，没有特定方向（适合单声道处理）
-      processorOptions: { chunkSize },
-    });
-    this._source.connect(this._workletNode);
-    this._workletNode.connect(this._audioContext.destination);
-    this._workletNode.port.onmessage = (event) => {
-      const { e, data } = event.data;
-      if (e === 'data') {
-        if (ignoreMute) {
-          if (Recorder.isSpeak(data, sampleRate)) {
-            this._pcmData.push(data);
-            this.ondataavailable?.(data);
-          }
-        } else {
-          this._pcmData.push(data);
-          this.ondataavailable?.(data);
-        }
-      }
-    };
-    this.onstart?.();
-  }
-  pause() {
-    this._workletNode?.port.postMessage('pause');
-    this.onpause?.();
-  }
-  resume() {
-    this._workletNode?.port.postMessage('resume');
-    this.onresume?.();
-  }
-  stop() {
-    this._workletNode?.port.postMessage('stop');
-    this._stream?.getTracks().forEach((track) => {
-      track.stop();
-    });
-    this._source?.disconnect();
-    this._workletNode?.disconnect();
-    this._stream = undefined;
-    this._source = undefined;
-    this._workletNode = undefined;
-    this._audioContext = undefined;
-    this.onstop?.(Recorder.pcmMerge(this._pcmData));
-    this._pcmData = [];
-  }
-  static pcmMerge(arrays) {
-    // 计算总长度
-    const totalLength = arrays.reduce((sum, arr) => sum + arr.length, 0);
-    // 创建合并后的 Int16Array
-    const merged = new Int16Array(totalLength);
-    let offset = 0;
-    for (const arr of arrays) {
-      // 将当前数组复制到目标数组的 offset 位置
-      merged.set(arr, offset);
-      offset += arr.length;
+    constructor(options) {
+        Object.defineProperty(this, "_stream", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "_audioContext", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "_source", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "_workletNode", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "_pcmData", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "_options", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "ondataavailable", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "onpause", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "onresume", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "onstart", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "onstop", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        this._options = options;
+        this._pcmData = [];
     }
-    return merged;
-  }
-  static pcmToWav(data, sampleRate) {
-    function writeString(view, offset, type) {
-      for (let i = 0; i < type.length; i++) {
-        view.setUint8(offset + i, type.charCodeAt(i));
-      }
+    async start() {
+        const { sampleRate, chunkSize, vad } = this._options;
+        this._stream = await navigator.mediaDevices.getUserMedia({
+            audio: true,
+        });
+        this._audioContext = new AudioContext({ sampleRate });
+        this._source = this._audioContext.createMediaStreamSource(this._stream);
+        await this._audioContext.audioWorklet.addModule(audioProcessorJs(), { credentials: 'include' });
+        this._workletNode = new AudioWorkletNode(this._audioContext, 'audio-processor', {
+            channelCount: 1, //单声道
+            channelCountMode: 'explicit', //输出通道数由 channelCount 显式指定
+            channelInterpretation: 'discrete', //通道是独立的音频流，没有特定方向（适合单声道处理）
+            processorOptions: { chunkSize },
+        });
+        this._source.connect(this._workletNode);
+        this._workletNode.connect(this._audioContext.destination);
+        this._workletNode.port.onmessage = (event) => {
+            const { e, data } = event.data;
+            if (e === 'data') {
+                if (vad) {
+                    if (Recorder.isSpeak(data, sampleRate, vad)) {
+                        this._pcmData.push(data);
+                        this.ondataavailable?.(data);
+                    }
+                }
+                else {
+                    this._pcmData.push(data);
+                    this.ondataavailable?.(data);
+                }
+            }
+        };
+        this.onstart?.();
     }
-    const numChannels = 1; // 单声道
-    const bytesPerSample = 2; // 16 位
-    const dataLength = data.length;
-    // 计算总文件大小
-    const fileSize = 44 + dataLength * bytesPerSample;
-    // 创建 ArrayBuffer
-    const buffer = new ArrayBuffer(fileSize);
-    const view = new DataView(buffer);
-    // 写入 RIFF 头
-    writeString(view, 0, 'RIFF');
-    view.setUint32(4, fileSize - 8, true); // 文件总长度 - 8
-    writeString(view, 8, 'WAVE');
-    // 写入 fmt 子块
-    writeString(view, 12, 'fmt ');
-    view.setUint32(16, 16, true); // 子块大小
-    view.setUint16(20, 1, true); // 音频格式 (PCM)
-    view.setUint16(22, numChannels, true); // 通道数
-    view.setUint32(24, sampleRate, true); // 采样率
-    view.setUint32(28, sampleRate * numChannels * bytesPerSample, true); // 字节率
-    view.setUint16(32, numChannels * bytesPerSample, true); // 块对齐
-    view.setUint16(34, bytesPerSample * 8, true); // 位深度 (16 位)
-    // 写入 data 子块
-    writeString(view, 36, 'data');
-    view.setUint32(40, dataLength * bytesPerSample, true); // 数据大小
-    // 写入 PCM 数据
-    for (let i = 0; i < dataLength; i++) {
-      view.setInt16(44 + i * bytesPerSample, data[i], true); // 小端序
+    pause() {
+        this._workletNode?.port.postMessage('pause');
+        this.onpause?.();
     }
-    const wavBlob = new Blob([buffer], { type: 'audio/wav' });
-    return wavBlob;
-  }
-  static isSpeak(int16Array, sampleRate) {
-    const frameSize = Math.floor(0.02 * sampleRate); // 20ms 帧长度
-    const threshold = 0.01; // 能量阈值（需根据实际调整）
-    let speakingFrames = 0;
-    const frameCount = Math.ceil(int16Array.length / frameSize);
-    for (let i = 0; i < frameCount; i++) {
-      const start = i * frameSize;
-      const end = Math.min(start + frameSize, int16Array.length);
-      const frame = int16Array.slice(start, end);
-      let sum = 0;
-      for (let j = 0; j < frame.length; j++) {
-        const sample = frame[j] / 32768; // 归一化到 [-1, 1]
-        sum += sample * sample;
-      }
-      const rms = Math.sqrt(sum / frame.length);
-      if (rms > threshold) {
-        speakingFrames++;
-      }
+    resume() {
+        this._workletNode?.port.postMessage('resume');
+        this.onresume?.();
     }
-    // 如果超过 50% 的帧有语音，认为在说话
-    return speakingFrames / frameCount > 0.5;
-  }
-  static parseWav(buffer) {
-    // 将32位无符号整数转换为4字节字符串
-    function uint32ToChars(uint32) {
-      return String.fromCharCode(
-        (uint32 >> 24) & 0xff,
-        (uint32 >> 16) & 0xff,
-        (uint32 >> 8) & 0xff,
-        (uint32 >> 0) & 0xff,
-      );
+    stop() {
+        this._workletNode?.port.postMessage('stop');
+        this._stream?.getTracks().forEach((track) => {
+            track.stop();
+        });
+        this._source?.disconnect();
+        this._workletNode?.disconnect();
+        this._stream = undefined;
+        this._source = undefined;
+        this._workletNode = undefined;
+        this._audioContext = undefined;
+        this.onstop?.(Recorder.pcmMerge(this._pcmData));
+        this._pcmData = [];
     }
-    const dataView = new DataView(buffer);
-    let offset = 0;
-    // 1. 验证RIFF头
-    const riff = dataView.getUint32(offset, false); // 大端序
-    if (riff !== 0x52494646) {
-      // 'RIFF' in hex
-      throw new Error('Not a RIFF file');
+    static pcmMerge(arrays) {
+        // 计算总长度
+        const totalLength = arrays.reduce((sum, arr) => sum + arr.length, 0);
+        // 创建合并后的 Int16Array
+        const merged = new Int16Array(totalLength);
+        let offset = 0;
+        for (const arr of arrays) {
+            // 将当前数组复制到目标数组的 offset 位置
+            merged.set(arr, offset);
+            offset += arr.length;
+        }
+        return merged;
     }
-    offset += 4;
-    const fileSize = dataView.getUint32(offset, true); // 小端序
-    offset += 4;
-    const wave = dataView.getUint32(offset, false); // 大端序
-    if (wave !== 0x57415645) {
-      // 'WAVE' in hex
-      throw new Error('Not a WAVE file');
+    static pcmToWav(data, sampleRate) {
+        function writeString(view, offset, type) {
+            for (let i = 0; i < type.length; i++) {
+                view.setUint8(offset + i, type.charCodeAt(i));
+            }
+        }
+        const numChannels = 1; // 单声道
+        const bytesPerSample = 2; // 16 位
+        const dataLength = data.length;
+        // 计算总文件大小
+        const fileSize = 44 + dataLength * bytesPerSample;
+        // 创建 ArrayBuffer
+        const buffer = new ArrayBuffer(fileSize);
+        const view = new DataView(buffer);
+        // 写入 RIFF 头
+        writeString(view, 0, 'RIFF');
+        view.setUint32(4, fileSize - 8, true); // 文件总长度 - 8
+        writeString(view, 8, 'WAVE');
+        // 写入 fmt 子块
+        writeString(view, 12, 'fmt ');
+        view.setUint32(16, 16, true); // 子块大小
+        view.setUint16(20, 1, true); // 音频格式 (PCM)
+        view.setUint16(22, numChannels, true); // 通道数
+        view.setUint32(24, sampleRate, true); // 采样率
+        view.setUint32(28, sampleRate * numChannels * bytesPerSample, true); // 字节率
+        view.setUint16(32, numChannels * bytesPerSample, true); // 块对齐
+        view.setUint16(34, bytesPerSample * 8, true); // 位深度 (16 位)
+        // 写入 data 子块
+        writeString(view, 36, 'data');
+        view.setUint32(40, dataLength * bytesPerSample, true); // 数据大小
+        // 写入 PCM 数据
+        for (let i = 0; i < dataLength; i++) {
+            view.setInt16(44 + i * bytesPerSample, data[i], true); // 小端序
+        }
+        const wavBlob = new Blob([buffer], { type: 'audio/wav' });
+        return wavBlob;
     }
-    offset += 4;
-    // 2. 遍历块，查找'fmt '块
-    while (offset < buffer.byteLength) {
-      const chunkId = dataView.getUint32(offset, false); // 大端序
-      const chunkSize = dataView.getUint32(offset + 4, true); // 小端序
-      const chunkIdStr = uint32ToChars(chunkId);
-      if (chunkIdStr === 'fmt ') {
-        // 3. 解析'fmt '块数据
-        const audioFormat = dataView.getUint16(offset + 8, true); // 小端序
-        const numChannels = dataView.getUint16(offset + 10, true);
-        const sampleRate = dataView.getUint32(offset + 12, true);
-        const byteRate = dataView.getUint32(offset + 16, true);
-        return { numChannels, sampleRate, byteRate, fileSize, audioFormat };
-      }
-      // 移动到下一个块
-      offset += 8 + chunkSize;
+    /**
+     * 语音活动检测（VAD）：判断音频片段是否包含说话内容
+     * @param int16Array 原始PCM音频数据（16位整型）
+     * @param sampleRate 采样率（如16000、44100）
+     * @param vad 场景类型或自定义配置
+     * @returns 是否检测到语音
+     */
+    static isSpeak(int16Array, sampleRate, vad) {
+        // 解析配置：场景预设优先，自定义配置可覆盖
+        let options;
+        if (typeof vad === 'string') {
+            options = { ...SCENE_PRESETS[vad] };
+        }
+        else {
+            // 默认使用普通室内场景配置，再合并自定义选项
+            options = { ...SCENE_PRESETS['normal-indoor'], ...vad };
+        }
+        // 计算帧长度（处理自定义帧时长）
+        const frameSize = Math.floor(((options.frameDurationMs ?? 20) * sampleRate) / 1000);
+        const totalFrameCount = Math.ceil(int16Array.length / frameSize);
+        let speakingFrameCount = 0;
+        let noiseRmsSum = 0;
+        // --------------------------
+        // 1. 计算背景噪声基准RMS
+        // --------------------------
+        const actualNoiseFrameCount = Math.min(options.noiseFrameCount, totalFrameCount);
+        for (let i = 0; i < actualNoiseFrameCount; i++) {
+            const start = i * frameSize;
+            const end = Math.min(start + frameSize, int16Array.length);
+            const frame = int16Array.subarray(start, end);
+            let sum = 0;
+            for (let j = 0; j < frame.length; j++) {
+                const sample = frame[j] / 32768; // 归一化到[-1, 1]
+                sum += sample * sample;
+            }
+            noiseRmsSum += Math.sqrt(sum / frame.length);
+        }
+        const baseNoiseRms = noiseRmsSum / actualNoiseFrameCount;
+        const dynamicThreshold = baseNoiseRms * options.noiseCoefficient;
+        // --------------------------
+        // 2. 逐帧检测语音
+        // --------------------------
+        for (let i = 0; i < totalFrameCount; i++) {
+            const start = i * frameSize;
+            const end = Math.min(start + frameSize, int16Array.length);
+            const frame = int16Array.subarray(start, end);
+            let sum = 0;
+            for (let j = 0; j < frame.length; j++) {
+                const sample = frame[j] / 32768;
+                sum += sample * sample;
+            }
+            const frameRms = Math.sqrt(sum / frame.length);
+            if (frameRms > dynamicThreshold) {
+                speakingFrameCount++;
+            }
+        }
+        // 3. 判断是否超过语音占比阈值
+        return speakingFrameCount / totalFrameCount > options.voiceRatioThreshold;
+    }
+    static parseWav(buffer) {
+        // 将32位无符号整数转换为4字节字符串
+        function uint32ToChars(uint32) {
+            return String.fromCharCode((uint32 >> 24) & 0xff, (uint32 >> 16) & 0xff, (uint32 >> 8) & 0xff, (uint32 >> 0) & 0xff);
+        }
+        const dataView = new DataView(buffer);
+        let offset = 0;
+        // 1. 验证RIFF头
+        const riff = dataView.getUint32(offset, false); // 大端序
+        if (riff !== 0x52494646) {
+            // 'RIFF' in hex
+            throw new Error('Not a RIFF file');
+        }
+        offset += 4;
+        const fileSize = dataView.getUint32(offset, true); // 小端序
+        offset += 4;
+        const wave = dataView.getUint32(offset, false); // 大端序
+        if (wave !== 0x57415645) {
+            // 'WAVE' in hex
+            throw new Error('Not a WAVE file');
+        }
+        offset += 4;
+        // 2. 遍历块，查找'fmt '块
+        while (offset < buffer.byteLength) {
+            const chunkId = dataView.getUint32(offset, false); // 大端序
+            const chunkSize = dataView.getUint32(offset + 4, true); // 小端序
+            const chunkIdStr = uint32ToChars(chunkId);
+            if (chunkIdStr === 'fmt ') {
+                // 3. 解析'fmt '块数据
+                const audioFormat = dataView.getUint16(offset + 8, true); // 小端序
+                const numChannels = dataView.getUint16(offset + 10, true);
+                const sampleRate = dataView.getUint32(offset + 12, true);
+                const byteRate = dataView.getUint32(offset + 16, true);
+                return { numChannels, sampleRate, byteRate, fileSize, audioFormat };
+            }
+            // 移动到下一个块
+            offset += 8 + chunkSize;
+        }
+        throw new Error('fmt chunk not found');
     }
-    throw new Error('fmt chunk not found');
-  }
 }
 
 export { Recorder };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "recorder-client",
-  "version": "1.0.6",
+  "version": "1.1.0",
   "description": "audio recording client",
   "type": "module",
   "main": "dist/index.cjs.js",
@@ -18,8 +18,6 @@
     "./*": "./*"
   },
   "scripts": {
-    "test": "vitest",
-    "eslint": "eslint --fix **/*.{ts,js}",
     "build": "rollup -c rollup.config.js"
   },
   "devDependencies": {
@@ -29,19 +27,18 @@
     "@rollup/plugin-replace": "^6.0.3",
     "@rollup/plugin-terser": "^0.4.4",
     "@rollup/plugin-typescript": "^12.3.0",
-    "@types/node": "^24.9.2",
-    "happy-dom": "^20.0.10",
-    "rollup": "^4.52.5",
-    "rollup-plugin-dts": "^6.2.3",
+    "@types/node": "^24.10.7",
+    "happy-dom": "^20.1.0",
+    "rollup": "^4.55.1",
+    "rollup-plugin-dts": "^6.3.0",
     "tslib": "^2.8.1",
     "typescript": "^5.9.3",
-    "typescript-eslint-standard": "^9.66.0",
-    "vitest": "^4.0.6"
+    "typescript-eslint-standard": "^9.75.0",
+    "vitest": "^4.0.16"
   },
   "author": "mivui",
   "license": "MIT",
   "keywords": [
-    "recorder",
     "audio",
     "MediaStream",
     "audioWorklet",