@vunk/plus 1.7.2 → 1.7.4

package/shared/audioToFrames/src/audio/StreamingFeatureExtractorService.d.ts

@@ -8,6 +8,18 @@
  *
  * @author Zhonghan Li
  */
+/**
+ * 创建AudioContext的兼容函数，处理Safari的限制。
+ * 在Safari中，AudioContext需要在用户交互后才能创建。
+ */
+export declare function createAudioContext(): Promise<AudioContext>;
+/**
+ * 特征提取结果接口
+ */
+export interface FeatureExtractionResult {
+    features: Float32Array;
+    dimensions: number[];
+}
 /**
  * 单个chunk的特征提取结果结构体。
  */
@@ -57,8 +69,6 @@ export declare class StreamingFeatureExtractorService {
     private queue;
     /** 回调函数集合 */
     private callbacks;
-    /** 性能计时器 */
-    private timer;
     /** 存储每个chunk的处理结果，用于最终合并 */
     private chunkResults;
     /** 音频总chunk数 */
@@ -73,8 +83,8 @@ export declare class StreamingFeatureExtractorService {
     private audioBuffer;
     /** 音频上下文实例，用于创建AudioBuffer */
     private audioContext;
-    /** 复用的特征提取器实例，以提高性能 */
-    private featureExtractor;
+    /** 复用的Worker实例，以提高性能 */
+    private worker;
     /**
      * 创建AudioBuffer的兼容方法，处理Safari的限制。
      * @param numberOfChannels 声道数
@@ -83,6 +93,12 @@ export declare class StreamingFeatureExtractorService {
      * @returns AudioBuffer实例
      */
     private createAudioBuffer;
+    /**
+     * 使用Worker处理AudioBuffer，提取特征
+     * @param audioBuffer 输入的AudioBuffer
+     * @returns Promise<FeatureExtractionResult>
+     */
+    private processWithWorker;
     /**
      * 启动流式音频特征提取。
      * @param audioBuffer 输入的音频缓冲区。

package/shared/audioToFrames/src/audio/index.d.ts

@@ -2,6 +2,4 @@
  * @file 音频处理服务模块导出
  * @author Zhonghan Li
  */
-export * from './AudioProcessor';
-export * from './FeatureExtractorService';
 export * from './StreamingFeatureExtractorService';

package/shared/audioToFrames/src/audio/index.mjs

@@ -1,111 +1,5 @@
-import { W as WINDOW_NUM_CHUNKS, d as WINDOW_NUM_CHUNKS_HALF, C as CHUNK_DURATION_SECONDS, a as NUM_SEQUENCE_FRAMES, N as NUM_MEL_BINS } from '../../../constants.mjs';
 import { w as workerConfig } from '../../../config.mjs';
-import { c as createErrorHandler, P as PerformanceTimer } from '../../../utils.mjs';
-
-function getAudioWindow(allAudioFeatures, allAudioDims, centerChunkIndex) {
-  if (allAudioDims.length !== 3) {
-    throw new Error(
-      "Invalid dimensions array provided to getAudioWindow. Expected length 3."
-    );
-  }
-  const [numChunks, chunkSize, melBins] = allAudioDims;
-  const finalWindowData = new Float32Array(
-    WINDOW_NUM_CHUNKS * chunkSize * melBins
-  );
-  const chunkTotalSize = chunkSize * melBins;
-  for (let j = 0; j < WINDOW_NUM_CHUNKS; j++) {
-    const sourceChunkIndex = centerChunkIndex - WINDOW_NUM_CHUNKS_HALF + j;
-    const clampedSourceIndex = Math.max(
-      0,
-      Math.min(sourceChunkIndex, numChunks - 1)
-    );
-    const sourceChunkStart = clampedSourceIndex * chunkTotalSize;
-    const sourceChunkEnd = sourceChunkStart + chunkTotalSize;
-    const chunkData = allAudioFeatures.subarray(
-      sourceChunkStart,
-      sourceChunkEnd
-    );
-    const destChunkStart = j * chunkTotalSize;
-    finalWindowData.set(chunkData, destChunkStart);
-  }
-  return finalWindowData;
-}
-async function createAudioContext() {
-  const AudioContextClass = window.AudioContext || window.webkitAudioContext;
-  if (!AudioContextClass) {
-    throw new Error("\u60A8\u7684\u6D4F\u89C8\u5668\u4E0D\u652F\u6301Web Audio API");
-  }
-  const audioContext = new AudioContextClass();
-  if (audioContext.state === "suspended") {
-    await audioContext.resume();
-  }
-  return audioContext;
-}
-
-class FeatureExtractorService {
-  constructor() {
-    this.errorHandler = createErrorHandler("FeatureExtractorService");
-    this.worker = new Worker(
-      /* @vite-ignore */
-      `${workerConfig.path}/feature.worker.js`,
-      { type: "module" }
-    );
-  }
-  /**
-   * 处理 AudioBuffer，在后台线程中提取特征。
-   * @param audioBuffer Web Audio API 的 AudioBuffer 对象。
-   * @returns 一个 Promise，成功时解析为特征结果，失败时拒绝并返回错误信息。
-   */
-  async process(audioBuffer) {
-    try {
-      const leftChannel = audioBuffer.getChannelData(0);
-      const rightChannel = audioBuffer.numberOfChannels > 1 ? audioBuffer.getChannelData(1) : void 0;
-      const message = {
-        leftChannel,
-        rightChannel,
-        sampleRate: audioBuffer.sampleRate
-      };
-      const transferList = [leftChannel.buffer, rightChannel?.buffer].filter(Boolean);
-      const result = await this.createWorkerPromise(
-        message,
-        transferList
-      );
-      return result;
-    } catch (error) {
-      throw this.errorHandler(error);
-    }
-  }
-  /**
-   * 创建Worker Promise，直接处理消息传递
-   */
-  createWorkerPromise(message, transferList) {
-    return new Promise((resolve, reject) => {
-      const handleMessage = (event) => {
-        this.worker.removeEventListener("message", handleMessage);
-        this.worker.removeEventListener("error", handleError);
-        if (event.data.status === "success") {
-          resolve(event.data.payload);
-        } else {
-          reject(new Error(event.data.error || "Unknown worker error"));
-        }
-      };
-      function handleError(error) {
-        this.worker.removeEventListener("message", handleMessage);
-        this.worker.removeEventListener("error", handleError);
-        reject(error);
-      }
-      this.worker.addEventListener("message", handleMessage);
-      this.worker.addEventListener("error", handleError);
-      this.worker.postMessage(message, transferList);
-    });
-  }
-  /**
-   * 终止Worker
-   */
-  terminate() {
-    this.worker.terminate();
-  }
-}
+import { C as CHUNK_DURATION_SECONDS, a as NUM_SEQUENCE_FRAMES, N as NUM_MEL_BINS } from '../../../constants.mjs';
 
 var __defProp = Object.defineProperty;
 var __defNormalProp = (obj, key, value) => key in obj ? __defProp(obj, key, { enumerable: true, configurable: true, writable: true, value }) : obj[key] = value;
@@ -239,13 +133,22 @@ var _AsyncQueue = class _AsyncQueue {
 __name(_AsyncQueue, "AsyncQueue");
 var AsyncQueue = _AsyncQueue;
 
+async function createAudioContext() {
+  const AudioContextClass = window.AudioContext || window.webkitAudioContext;
+  if (!AudioContextClass) {
+    throw new Error("\u60A8\u7684\u6D4F\u89C8\u5668\u4E0D\u652F\u6301Web Audio API");
+  }
+  const audioContext = new AudioContextClass();
+  if (audioContext.state === "suspended") {
+    await audioContext.resume();
+  }
+  return audioContext;
+}
 class StreamingFeatureExtractorService {
   constructor() {
     this.queue = new AsyncQueue();
     /** 回调函数集合 */
     this.callbacks = {};
-    /** 性能计时器 */
-    this.timer = new PerformanceTimer();
     /** 存储每个chunk的处理结果，用于最终合并 */
     this.chunkResults = [];
     /** 音频总chunk数 */
@@ -260,8 +163,8 @@ class StreamingFeatureExtractorService {
     this.audioBuffer = null;
     /** 音频上下文实例，用于创建AudioBuffer */
     this.audioContext = null;
-    /** 复用的特征提取器实例，以提高性能 */
-    this.featureExtractor = null;
+    /** 复用的Worker实例，以提高性能 */
+    this.worker = null;
   }
   /**
    * 创建AudioBuffer的兼容方法，处理Safari的限制。
@@ -289,6 +192,47 @@ class StreamingFeatureExtractorService {
       );
     }
   }
+  /**
+   * 使用Worker处理AudioBuffer，提取特征
+   * @param audioBuffer 输入的AudioBuffer
+   * @returns Promise<FeatureExtractionResult>
+   */
+  async processWithWorker(audioBuffer) {
+    if (!this.worker) {
+      throw new Error("Worker not initialized");
+    }
+    return new Promise((resolve, reject) => {
+      const cleanup = () => {
+        if (this.worker) {
+          this.worker.removeEventListener("message", handleMessage);
+          this.worker.removeEventListener("error", handleError);
+        }
+      };
+      function handleMessage(event) {
+        cleanup();
+        if (event.data.status === "success") {
+          resolve(event.data.payload);
+        } else {
+          reject(new Error(event.data.error || "Unknown worker error"));
+        }
+      }
+      function handleError(error) {
+        cleanup();
+        reject(error);
+      }
+      this.worker.addEventListener("message", handleMessage);
+      this.worker.addEventListener("error", handleError);
+      const leftChannel = audioBuffer.getChannelData(0);
+      const rightChannel = audioBuffer.numberOfChannels > 1 ? audioBuffer.getChannelData(1) : void 0;
+      const message = {
+        leftChannel,
+        rightChannel,
+        sampleRate: audioBuffer.sampleRate
+      };
+      const transferList = [leftChannel.buffer, rightChannel?.buffer].filter(Boolean);
+      this.worker.postMessage(message, transferList);
+    });
+  }
   /**
    * 启动流式音频特征提取。
    * @param audioBuffer 输入的音频缓冲区。
@@ -305,22 +249,24 @@ class StreamingFeatureExtractorService {
     this.chunkResults = [];
     this.completedChunks = 0;
     this.totalFramesProcessed = 0;
-    this.featureExtractor = new FeatureExtractorService();
+    if (!this.worker) {
+      this.worker = new Worker(
+        /* @vite-ignore */
+        `${workerConfig.path}/feature.worker.js`,
+        { type: "module" }
+      );
+    }
     const durationSeconds = audioBuffer.duration;
     this.totalChunks = Math.ceil(durationSeconds / CHUNK_DURATION_SECONDS);
     try {
-      this.timer.start("totalProcessing");
       for (let chunkIndex = 0; chunkIndex < this.totalChunks; chunkIndex++) {
         if (!this.isRunning)
           break;
-        this.timer.start(`chunk_${chunkIndex}`);
         await this.processChunk(chunkIndex);
-        this.timer.end(`chunk_${chunkIndex}`);
       }
       if (this.isRunning) {
         await this.finalize();
       }
-      this.timer.end("totalProcessing");
     } catch (error) {
       const errorMessage = error instanceof Error ? error.message : "Unknown error";
       this.callbacks.onError?.(errorMessage);
@@ -334,7 +280,7 @@ class StreamingFeatureExtractorService {
    * @param chunkIndex chunk的索引（从0开始）。
    */
   async processChunk(chunkIndex) {
-    if (!this.audioBuffer || !this.featureExtractor)
+    if (!this.audioBuffer || !this.worker)
       return;
     const startTimeSeconds = chunkIndex * CHUNK_DURATION_SECONDS;
     const endTimeSeconds = Math.min(
@@ -346,11 +292,6 @@ class StreamingFeatureExtractorService {
     );
     const endSample = Math.floor(endTimeSeconds * this.audioBuffer.sampleRate);
     const chunkLength = endSample - startSample;
-    console.log(
-      `\u5904\u7406chunk ${chunkIndex}: ${startTimeSeconds.toFixed(
-        2
-      )}s - ${endTimeSeconds.toFixed(2)}s`
-    );
     const chunkBuffer = await this.createAudioBuffer(
       this.audioBuffer.numberOfChannels,
       chunkLength,
@@ -361,36 +302,25 @@ class StreamingFeatureExtractorService {
       const chunkData = chunkBuffer.getChannelData(channel);
       chunkData.set(sourceData.subarray(startSample, endSample));
     }
-    try {
-      this.timer.start(`extract_chunk_${chunkIndex}`);
-      const result = await this.featureExtractor.process(chunkBuffer);
-      this.timer.end(`extract_chunk_${chunkIndex}`);
-      const chunkResult = {
-        chunkIndex,
-        features: result.features,
-        dimensions: result.dimensions,
-        startTimeSeconds,
-        endTimeSeconds
-      };
-      this.totalFramesProcessed += chunkResult.dimensions[0];
-      this.chunkResults[chunkIndex] = chunkResult;
-      this.completedChunks++;
-      console.log(
-        `Chunk ${chunkIndex} \u5B8C\u6210\uFF0C\u7279\u5F81\u7EF4\u5EA6: [${result.dimensions.join(", ")}]`
-      );
-      this.callbacks.onChunkComplete?.(chunkResult);
-      this.callbacks.onProgress?.(this.completedChunks, this.totalChunks);
-      this.chunkResults[chunkIndex] = null;
-    } catch (error) {
-      console.error(`\u5904\u7406chunk ${chunkIndex} \u65F6\u53D1\u751F\u9519\u8BEF:`, error);
-      throw error;
-    }
+    const result = await this.processWithWorker(chunkBuffer);
+    const chunkResult = {
+      chunkIndex,
+      features: result.features,
+      dimensions: result.dimensions,
+      startTimeSeconds,
+      endTimeSeconds
+    };
+    this.totalFramesProcessed += chunkResult.dimensions[0];
+    this.chunkResults[chunkIndex] = chunkResult;
+    this.completedChunks++;
+    this.callbacks.onChunkComplete?.(chunkResult);
+    this.callbacks.onProgress?.(this.completedChunks, this.totalChunks);
+    this.chunkResults[chunkIndex] = null;
   }
   /**
    * 所有chunk处理完成后，合并特征并调用onComplete。
    */
   async finalize() {
-    console.log("\u6700\u7EC8\u786E\u5B9A\u6240\u6709chunk\u7684\u5904\u7406...");
     if (this.completedChunks === 0) {
       this.callbacks.onError?.("No chunks were processed.");
       return;
@@ -400,7 +330,6 @@ class StreamingFeatureExtractorService {
       NUM_SEQUENCE_FRAMES,
       NUM_MEL_BINS
     ];
-    console.log(`\u7279\u5F81\u5904\u7406\u5B8C\u6210, \u603B\u7EF4\u5EA6: [${totalDimensions.join(", ")}]`);
     this.callbacks.onComplete?.(totalDimensions);
   }
   /**
@@ -409,9 +338,9 @@ class StreamingFeatureExtractorService {
   cleanup() {
     this.isRunning = false;
     this.audioBuffer = null;
-    if (this.featureExtractor) {
-      this.featureExtractor.terminate();
-      this.featureExtractor = null;
+    if (this.worker) {
+      this.worker.terminate();
+      this.worker = null;
     }
     this.chunkResults.forEach((chunk) => {
       if (chunk && chunk.features) {
@@ -423,18 +352,16 @@ class StreamingFeatureExtractorService {
     this.completedChunks = 0;
     this.totalChunks = 0;
     this.totalFramesProcessed = 0;
-    this.timer.reset();
   }
   /**
    * 停止当前的流式处理。
    */
   stop() {
     if (this.isRunning) {
-      console.log("\u505C\u6B62\u6D41\u5F0F\u97F3\u9891\u5904\u7406");
       this.isRunning = false;
-      if (this.featureExtractor) {
-        this.featureExtractor.terminate();
-        this.featureExtractor = null;
+      if (this.worker) {
+        this.worker.terminate();
+        this.worker = null;
       }
       this.callbacks = {};
       this.chunkResults = [];
@@ -445,4 +372,4 @@ class StreamingFeatureExtractorService {
   }
 }
 
-export { FeatureExtractorService, NUM_MEL_BINS, NUM_SEQUENCE_FRAMES, StreamingFeatureExtractorService, createAudioContext, getAudioWindow };
+export { StreamingFeatureExtractorService, createAudioContext };

package/shared/audioToFrames/src/core/constants.d.ts

@@ -11,29 +11,7 @@ export declare const NUM_MEL_BINS = 80;
 export declare const NUM_SEQUENCE_FRAMES = 4;
 /** 每个chunk的时长（秒） */
 export declare const CHUNK_DURATION_SECONDS = 5;
-/** 目标帧率 (FPS) */
-export declare const TARGET_FPS = 25;
-/** 每帧之间的间隔时间 (毫秒) */
-export declare const FRAME_INTERVAL_MS: number;
-/** 音频采样率 (Hz) */
-export declare const SAMPLE_RATE = 16000;
-/** 音频帧长度 (毫秒) */
-export declare const FRAME_LENGTH_MS = 25;
-/** 音频帧移 (毫秒) */
-export declare const FRAME_SHIFT_MS = 10;
-/** 能量下限，用于防止在对数计算中出现 log(0) 的情况 */
-export declare const ENERGY_FLOOR = 1e-10;
-/** 预加重系数 */
-export declare const PREEMPH_COEFF = 0.97;
 /** 上下文窗口中的总块数 */
 export declare const WINDOW_NUM_CHUNKS = 32;
 /** 上下文窗口的单边块数（半径） */
 export declare const WINDOW_NUM_CHUNKS_HALF: number;
-/** 批处理大小，用于批量推理以提高吞吐量 */
-export declare const BATCH_SIZE = 32;
-/** 默认的低频截止频率 (Hz) */
-export declare const DEFAULT_LOW_FREQ = 20;
-/** 是否移除直流分量（DC offset） */
-export declare const REMOVE_DC_OFFSET = true;
-/** FFT 计算大小是否向上舍入到最近的2的幂 */
-export declare const ROUND_TO_POWER_OF_TWO = true;