Add Flutter example for speaker identification (#894)

sherpa-onnx/flutter/example/lib/speaker_identification_test.dart

@@ -0,0 +1,129 @@
+// Copyright (c)  2024  Xiaomi Corporation
+import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
+import 'dart:typed_data';
+import 'package:path/path.dart';
+import './utils.dart';
+
+Float32List computeEmbedding(
+    {required sherpa_onnx.SpeakerEmbeddingExtractor extractor,
+    required String filename}) {
+  final stream = extractor.createStream();
+  final waveData = sherpa_onnx.readWave(filename);
+
+  stream.acceptWaveform(
+      samples: waveData.samples, sampleRate: waveData.sampleRate);
+
+  stream.inputFinished();
+
+  final embedding = extractor.compute(stream);
+
+  stream.free();
+
+  return embedding;
+}
+
+Future<void> testSpeakerID() async {
+  final src =
+      'assets/3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx';
+  final modelPath = await copyAssetFile(src: src, dst: 'model.onnx');
+
+  final config = sherpa_onnx.SpeakerEmbeddingExtractorConfig(model: modelPath);
+  final extractor = sherpa_onnx.SpeakerEmbeddingExtractor(config: config);
+
+  const _spk1Files = [
+    'assets/sr-data/enroll/fangjun-sr-1.wav',
+    'assets/sr-data/enroll/fangjun-sr-2.wav',
+    'assets/sr-data/enroll/fangjun-sr-3.wav',
+  ];
+  final spk1Files = <String>[];
+  for (final f in _spk1Files) {
+    spk1Files.add(await copyAssetFile(src: f, dst: basename(f)));
+  }
+
+  final spk1Vec = <Float32List>[];
+  for (final f in spk1Files) {
+    spk1Vec.add(computeEmbedding(extractor: extractor, filename: f));
+  }
+
+  const _spk2Files = [
+    'assets/sr-data/enroll/leijun-sr-1.wav',
+    'assets/sr-data/enroll/leijun-sr-2.wav',
+  ];
+  final spk2Files = <String>[];
+  for (final f in _spk2Files) {
+    spk2Files.add(await copyAssetFile(src: f, dst: basename(f)));
+  }
+
+  final spk2Vec = <Float32List>[];
+  for (final f in spk2Files) {
+    spk2Vec.add(computeEmbedding(extractor: extractor, filename: f));
+  }
+
+  final manager = sherpa_onnx.SpeakerEmbeddingManager(extractor.dim);
+  assert(manager.numSpeakers == 0, '${manager.numSpeakers}');
+
+  bool ok = manager.addMulti(name: 'fangjun', embeddingList: spk1Vec);
+  assert(ok, "Failed to add fangjun");
+  assert(manager.numSpeakers == 1, '${manager.numSpeakers}');
+
+  ok = manager.addMulti(name: 'leijun', embeddingList: spk2Vec);
+  assert(ok, "Failed to add leijun");
+  assert(manager.numSpeakers == 2, '${manager.numSpeakers}');
+
+  bool found = manager.contains('fangjun');
+  assert(found, 'Failed to find fangjun');
+
+  found = manager.contains('leijun');
+  assert(found, 'Failed to find leijun');
+
+  print('---All speakers---');
+
+  print(manager.allSpeakerNames);
+
+  print('------------');
+
+  const _testFiles = [
+    'assets/sr-data/test/fangjun-test-sr-1.wav',
+    'assets/sr-data/test/leijun-test-sr-1.wav',
+    'assets/sr-data/test/liudehua-test-sr-1.wav',
+  ];
+
+  final testFiles = <String>[];
+  for (final f in _testFiles) {
+    testFiles.add(await copyAssetFile(src: f, dst: basename(f)));
+  }
+
+  const threshold = 0.6;
+
+  for (final f in testFiles) {
+    final embedding = computeEmbedding(extractor: extractor, filename: f);
+
+    var name = manager.search(embedding: embedding, threshold: threshold);
+    if (name == '') {
+      name = '<Unknown>';
+    }
+    print('${f}: ${name}');
+  }
+
+  ok = manager.verify(
+      name: 'fangjun',
+      embedding: computeEmbedding(extractor: extractor, filename: testFiles[0]),
+      threshold: threshold);
+  assert(ok, 'Failed to verify fangjun using ${testFiles[0]}');
+
+  ok = manager.remove('fangjun');
+  assert(ok, 'Failed to remove fangjun');
+  assert(manager.numSpeakers == 1, '${manager.numSpeakers}');
+
+  found = manager.contains('fangjun');
+  assert(!found, 'Still found fangjun!');
+
+  ok = manager.verify(
+      name: 'fangjun',
+      embedding: computeEmbedding(extractor: extractor, filename: testFiles[0]),
+      threshold: threshold);
+  assert(!ok, '${testFiles[0]}');
+
+  manager.free();
+  extractor.free();
+}