First working version for decoding a file

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/ets/components/NonStreamingSpeakerDiarization.ets

@@ -32,8 +32,9 @@ export class OfflineSpeakerDiarizationConfig {
 }
 
 export class OfflineSpeakerDiarizationSegment {
-  public start: number = 0; // in secondspublic end: number = 0; // in secondspublic speaker: number =
-    0; // ID of the speaker; count from 0
+  public start: number = 0; // in seconds
+  public end: number = 0; // in seconds
+  public speaker: number = 0; // ID of the speaker; count from 0
 }
 
 export class OfflineSpeakerDiarization {
@@ -62,8 +63,8 @@ export class OfflineSpeakerDiarization {
    *    "speaker": an_integer,
    *  }
    */
-  process(samples: Float32Array): OfflineSpeakerDiarizationSegment {
-    return offlineSpeakerDiarizationProcess(this.handle, samples) as OfflineSpeakerDiarizationSegment;
+  process(samples: Float32Array): OfflineSpeakerDiarizationSegment[] {
+    return offlineSpeakerDiarizationProcess(this.handle, samples) as OfflineSpeakerDiarizationSegment[];
   }
 
   setConfig(config: OfflineSpeakerDiarizationConfig) {

harmony-os/SherpaOnnxSpeakerDiarization/entry/build-profile.json5

@@ -1,6 +1,11 @@
 {
   "apiType": "stageMode",
   "buildOption": {
+    "sourceOption": {
+      "workers": [
+        './src/main/ets/workers/SpeakerDiarizationWorker.ets'
+      ]
+    }
   },
   "buildOptionSet": [
     {

harmony-os/SherpaOnnxSpeakerDiarization/entry/src/main/ets/pages/Index.ets

@@ -1,4 +1,8 @@
 import { LengthUnit } from '@kit.ArkUI';
+import worker, { MessageEvents } from '@ohos.worker';
+import { systemTime, BusinessError } from '@kit.BasicServicesKit';
+import { picker } from '@kit.CoreFileKit';
+
 
 @Entry
 @Component
@@ -8,11 +12,17 @@ struct Index {
   @State currentIndex: number = 0;
   private controller: TabsController = new TabsController();
 
+  private workerInstance?: worker.ThreadWorker
+  private readonly scriptURL: string = 'entry/ets/workers/SpeakerDiarizationWorker.ets'
+
   @State resultForFile: string = '';
   @State resultForMic: string = '';
 
   @State micBtnCaption: string = 'Start recording';
-  @State micSaveBtnCaption: string = 'Save';
+  @State micSaveBtnCaption: string = 'Save audio';
+
+  @State micBtnEnabled: boolean = false;
+  @State selectFileBtnEnabled: boolean = false;
 
   @Builder
   TabBuilder(title: string, targetIndex: number, selectedImg: Resource, normalImg: Resource) {
@@ -25,13 +35,79 @@ struct Index {
     })
   }
 
+  aboutToAppear(): void {
+    this.workerInstance = new worker.ThreadWorker(this.scriptURL, {
+      name: 'Streaming ASR worker'
+    });
+
+    this.workerInstance.onmessage = (e: MessageEvents) => {
+      const msgType = e.data['msgType'] as string;
+      console.log(`received msg from worker: ${msgType}`);
+
+      if (msgType == 'init-speaker-diarization-done') {
+        console.log('Speaker diarization initialized successfully');
+
+        this.resultForFile = 'Initialization finished.\nPlease select a .wav file.';
+        this.resultForMic = 'Initialization finished.\nPlease click the button Start recording.';
+
+        this.selectFileBtnEnabled = true;
+        this.micBtnEnabled = true;
+      }
+
+      if (msgType == 'speaker-diarization-file-done') {
+        const result = e.data['result'] as string;
+        this.resultForFile = result;
+
+        this.selectFileBtnEnabled = true;
+      }
+    };
+
+    const context = getContext();
+    this.workerInstance.postMessage({ msgType: 'init-speaker-diarization', context });
+    console.log('initializing');
+    this.resultForFile = 'Initializing models. Please wait';
+    this.resultForMic = this.resultForFile;
+  }
+
   build() {
     Column() {
       Tabs({ barPosition: BarPosition.End, controller: this.controller }) {
         TabContent() {
           Column({space: 10}) {
             Text(this.title).fontSize(this.titleFontSize).fontWeight(FontWeight.Bold);
             Button('Select .wav file (16kHz) ')
+              .enabled(this.selectFileBtnEnabled)
+              .onClick(()=>{
+                this.resultForFile = '';
+                this.selectFileBtnEnabled = false;
+
+                const documentSelectOptions = new picker.DocumentSelectOptions();
+                documentSelectOptions.maxSelectNumber = 1;
+                documentSelectOptions.fileSuffixFilters = ['.wav'];
+                const documentViewPicker = new picker.DocumentViewPicker();
+
+                documentViewPicker.select(documentSelectOptions).then((result: Array<string>) => {
+                  console.log(`select file result: ${result}`);
+
+                  if (!result[0]) {
+                    this.resultForFile = 'Please select a file to decode';
+                    this.selectFileBtnEnabled = true;
+                    return;
+                  }
+
+                  if (this.workerInstance) {
+                    this.workerInstance.postMessage({
+                      msgType: 'speaker-diarization-file', filename: result[0],
+                    });
+                    this.resultForFile = `Decoding ${result[0]} ... ...`;
+                  } else {
+                    console.log(`this worker instance is undefined ${this.workerInstance}`);
+                  }
+                }).catch((err: BusinessError) => {
+                  console.error(`Failed to select file, code is ${err.code}, message is ${err.message}`);
+                  this.selectFileBtnEnabled = true;
+                })
+              })
 
             TextArea({ text: this.resultForFile })
               .lineSpacing({ value: 10, unit: LengthUnit.VP })

harmony-os/SherpaOnnxSpeakerDiarization/entry/src/main/ets/workers/SpeakerDiarizationWorker.ets

@@ -0,0 +1,103 @@
+import worker, { ThreadWorkerGlobalScope, MessageEvents, ErrorEvent } from '@ohos.worker';
+import { OfflineSpeakerDiarization, OfflineSpeakerDiarizationConfig,
+  OfflineSpeakerDiarizationSegment,
+  readWaveFromBinary, Samples } from 'sherpa_onnx';
+import { fileIo } from '@kit.CoreFileKit';
+
+const workerPort: ThreadWorkerGlobalScope = worker.workerPort;
+
+let sd: OfflineSpeakerDiarization;
+
+function readWave(filename: string): Samples {
+  const fp = fileIo.openSync(filename);
+  const stat = fileIo.statSync(fp.fd);
+  const arrayBuffer = new ArrayBuffer(stat.size);
+  fileIo.readSync(fp.fd, arrayBuffer);
+  const data: Uint8Array = new Uint8Array(arrayBuffer);
+  return readWaveFromBinary(data) as Samples;
+}
+
+function initOfflineSpeakerDiarization(context: Context): OfflineSpeakerDiarization {
+  const config: OfflineSpeakerDiarizationConfig = new OfflineSpeakerDiarizationConfig();
+
+  config.segmentation.pyannote.model = 'sherpa-onnx-pyannote-segmentation-3-0/model.int8.onnx';
+  config.segmentation.numThreads = 2;
+  config.segmentation.debug = true;
+
+  config.embedding.model = '3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx';
+  config.embedding.numThreads = 2;
+  config.embedding.debug = true;
+
+  config.minDurationOn = 0.2;
+  config.minDurationOff = 0.5;
+  return new OfflineSpeakerDiarization(config, context.resourceManager);
+}
+
+/**
+ * Defines the event handler to be called when the worker thread receives a message sent by the host thread.
+ * The event handler is executed in the worker thread.
+ *
+ * @param e message data
+ */
+workerPort.onmessage = (e: MessageEvents) => {
+  const msgType = e.data['msgType'] as string;
+
+  console.log(`from the main thread, msg-type: ${msgType}`);
+  if (msgType == 'init-speaker-diarization' && !sd) {
+    const context: Context = e.data['context'] as Context;
+    sd = initOfflineSpeakerDiarization(context);
+    workerPort.postMessage({msgType: 'init-speaker-diarization-done'});
+    console.log('Init sd done');
+  }
+
+  if (msgType == 'speaker-diarization-file') {
+    const filename = e.data['filename'] as string;
+    const wave = readWave(filename);
+    let result = '';
+    if (wave == undefined || wave == null) {
+      result = `Failed to read ${filename}`;
+    } else if (wave.sampleRate != sd.sampleRate) {
+      result = `Expected sample rate: ${sd.sampleRate}`;
+      result += '\n';
+      result += `Sample rate in file ${filename} is ${wave.sampleRate}`;
+    } else {
+      const duration = wave.samples.length / wave.sampleRate;
+      console.log(`Processing ${filename} of ${duration} seconds`);
+
+      const r: OfflineSpeakerDiarizationSegment[] = sd.process(wave.samples)
+      console.log(`r is ${r.length}, ${r}`);
+      for (const s of r) {
+        const start: string = s.start.toFixed(3);
+        const end: string = s.end.toFixed(3);
+        result += `${start}\t--\t${end}\tspeaker_${s.speaker}\n`;
+        console.log(`result: ${result}`);
+      }
+
+      if (r.length == 0) {
+        result = 'The result is empty';
+      }
+    }
+
+    workerPort.postMessage({
+      msgType: 'speaker-diarization-file-done',
+      result
+    });
+  }
+}
+/**
+ * Defines the event handler to be called when the worker receives a message that cannot be deserialized.
+ * The event handler is executed in the worker thread.
+ *
+ * @param e message data
+ */
+workerPort.onmessageerror = (e: MessageEvents) => {
+}
+
+/**
+ * Defines the event handler to be called when an exception occurs during worker execution.
+ * The event handler is executed in the worker thread.
+ *
+ * @param e error message
+ */
+workerPort.onerror = (e: ErrorEvent) => {
+}

harmony-os/SherpaOnnxSpeakerDiarization/entry/src/main/resources/base/element/string.json

@@ -2,15 +2,19 @@
   "string": [
     {
       "name": "module_desc",
-      "value": "module description"
+      "value": "On-device speaker diarization with Next-gen Kaldi"
     },
     {
       "name": "EntryAbility_desc",
-      "value": "description"
+      "value": "On-device speaker diarization with Next-gen Kaldi"
     },
     {
       "name": "EntryAbility_label",
-      "value": "label"
+      "value": "Speaker diarization"
+    },
+    {
+      "name": "mic_reason",
+      "value": "access the microphone for on-device speaker diarizatiaon with Next-gen Kaldi"
     }
   ]
 }

harmony-os/SherpaOnnxSpeakerDiarization/entry/src/main/resources/en_US/element/string.json

@@ -2,15 +2,19 @@
   "string": [
     {
       "name": "module_desc",
-      "value": "module description"
+      "value": "On-device speaker diarization with Next-gen Kaldi"
     },
     {
       "name": "EntryAbility_desc",
-      "value": "description"
+      "value": "On-device speaker diarization with Next-gen Kaldi"
     },
     {
       "name": "EntryAbility_label",
-      "value": "label"
+      "value": "Speaker diarization"
+    },
+    {
+      "name": "mic_reason",
+      "value": "access the microphone for on-device speaker diarizatiaon with Next-gen Kaldi"
     }
   ]
 }

harmony-os/SherpaOnnxSpeakerDiarization/entry/src/main/resources/zh_CN/element/string.json

@@ -2,15 +2,19 @@
   "string": [
     {
       "name": "module_desc",
-      "value": "模块描述"
+      "value": "新一代Kaldi: 本地说话人日志"
     },
     {
       "name": "EntryAbility_desc",
-      "value": "description"
+      "value": "新一代Kaldi: 本地说话人日志"
     },
     {
       "name": "EntryAbility_label",
-      "value": "label"
+      "value": "说话人日志"
+    },
+    {
+      "name": "mic_reason",
+      "value": "使用新一代Kaldi, 访问麦克风进行本地说话人日志 (不需要联网)"
     }
   ]
 }

sherpa-onnx/c-api/c-api.cc

@@ -2053,11 +2053,6 @@ SherpaOnnxCreateOfflineSpeakerDiarizationOHOS(
 
   auto sd_config = GetOfflineSpeakerDiarizationConfig(config);
 
-  if (!sd_config.Validate()) {
-    SHERPA_ONNX_LOGE("Errors in config");
-    return nullptr;
-  }
-
   SherpaOnnxOfflineSpeakerDiarization *sd =
       new SherpaOnnxOfflineSpeakerDiarization;