Add Speaker ID demo for C#

.github/scripts/test-dot-net.sh

@@ -2,7 +2,10 @@
 
 cd dotnet-examples/
 
-cd streaming-hlg-decoding/
+cd speaker-identification
+./run.sh
+
+cd ../streaming-hlg-decoding/
 ./run.sh
 
 cd ../spoken-language-identification

.github/workflows/test-dot-net.yaml

@@ -179,6 +179,7 @@ jobs:
           cp -v scripts/dotnet/examples/speech-recognition-from-microphone.csproj dotnet-examples/speech-recognition-from-microphone/
           cp -v scripts/dotnet/examples/spoken-language-identification.csproj dotnet-examples/spoken-language-identification/
           cp -v scripts/dotnet/examples/streaming-hlg-decoding.csproj dotnet-examples/streaming-hlg-decoding
+          cp -v scripts/dotnet/examples/speaker-identification.csproj dotnet-examples/speaker-identification
 
           ls -lh /tmp
 

CMakeLists.txt

@@ -1,7 +1,7 @@
 cmake_minimum_required(VERSION 3.13 FATAL_ERROR)
 project(sherpa-onnx)
 
-set(SHERPA_ONNX_VERSION "1.9.23")
+set(SHERPA_ONNX_VERSION "1.9.24")
 
 # Disable warning about
 #

dotnet-examples/sherpa-onnx.sln

@@ -17,6 +17,8 @@ Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "spoken-language-identificat
 EndProject
 Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "streaming-hlg-decoding", "streaming-hlg-decoding\streaming-hlg-decoding.csproj", "{C4A368A5-FCA0-419D-97C9-C8CE0B08EB99}"
 EndProject
+Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "speaker-identification", "speaker-identification\speaker-identification.csproj", "{2B1B140E-A92F-426B-B0DF-5D916B67304F}"
+EndProject
 Global
 	GlobalSection(SolutionConfigurationPlatforms) = preSolution
 		Debug|Any CPU = Debug|Any CPU
@@ -54,5 +56,9 @@ Global
 		{C4A368A5-FCA0-419D-97C9-C8CE0B08EB99}.Debug|Any CPU.Build.0 = Debug|Any CPU
 		{C4A368A5-FCA0-419D-97C9-C8CE0B08EB99}.Release|Any CPU.ActiveCfg = Release|Any CPU
 		{C4A368A5-FCA0-419D-97C9-C8CE0B08EB99}.Release|Any CPU.Build.0 = Release|Any CPU
+		{2B1B140E-A92F-426B-B0DF-5D916B67304F}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{2B1B140E-A92F-426B-B0DF-5D916B67304F}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{2B1B140E-A92F-426B-B0DF-5D916B67304F}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{2B1B140E-A92F-426B-B0DF-5D916B67304F}.Release|Any CPU.Build.0 = Release|Any CPU
 	EndGlobalSection
 EndGlobal

dotnet-examples/speaker-identification/Program.cs

@@ -0,0 +1,155 @@
+// Copyright (c)  2024  Xiaomi Corporation
+//
+// This file shows how to do speaker identification with sherpa-onnx.
+//
+// 1. Download a model from
+// https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-recongition-models
+//
+// wget https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-recongition-models/3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx
+//
+// 2. Download test data from
+//
+// git clone https://github.com/csukuangfj/sr-data
+//
+// 3. Now run it
+//
+// dotnet run
+
+using SherpaOnnx;
+using System.Collections.Generic;
+using System;
+
+class SpeakerIdentificationDemo
+{
+  public static float[] ComputeEmbedding(SpeakerEmbeddingExtractor extractor, String filename)
+  {
+    WaveReader reader = new WaveReader(filename);
+
+    OnlineStream stream = extractor.CreateStream();
+    stream.AcceptWaveform(reader.SampleRate, reader.Samples);
+    stream.InputFinished();
+
+    float[] embedding = extractor.Compute(stream);
+
+    return embedding;
+  }
+
+  static void Main(string[] args)
+  {
+    var config = new SpeakerEmbeddingExtractorConfig();
+    config.Model = "./3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx";
+    config.Debug = 1;
+    var extractor = new SpeakerEmbeddingExtractor(config);
+
+    var manager = new SpeakerEmbeddingManager(extractor.Dim);
+
+    string[] spk1Files =
+        new string[] {
+          "./sr-data/enroll/fangjun-sr-1.wav",
+          "./sr-data/enroll/fangjun-sr-2.wav",
+          "./sr-data/enroll/fangjun-sr-3.wav",
+        };
+    float[][] spk1Vec = new float[spk1Files.Length][];
+
+    for (int i = 0; i < spk1Files.Length; ++i)
+    {
+      spk1Vec[i] = ComputeEmbedding(extractor, spk1Files[i]);
+    }
+
+    string[] spk2Files =
+        new string[] {
+          "./sr-data/enroll/leijun-sr-1.wav", "./sr-data/enroll/leijun-sr-2.wav",
+        };
+
+    float[][] spk2Vec = new float[spk2Files.Length][];
+
+    for (int i = 0; i < spk2Files.Length; ++i)
+    {
+      spk2Vec[i] = ComputeEmbedding(extractor, spk2Files[i]);
+    }
+
+    if (!manager.Add("fangjun", spk1Vec))
+    {
+      Console.WriteLine("Failed to register fangjun");
+      return;
+    }
+
+    if (!manager.Add("leijun", spk2Vec))
+    {
+      Console.WriteLine("Failed to register leijun");
+      return;
+    }
+
+    if (manager.NumSpeakers != 2)
+    {
+      Console.WriteLine("There should be two speakers");
+      return;
+    }
+
+    if (!manager.Contains("fangjun"))
+    {
+      Console.WriteLine("It should contain the speaker fangjun");
+      return;
+    }
+
+    if (!manager.Contains("leijun"))
+    {
+      Console.WriteLine("It should contain the speaker leijun");
+      return;
+    }
+
+    Console.WriteLine("---All speakers---");
+
+    string[] allSpeakers = manager.GetAllSpeakers();
+    foreach (var s in allSpeakers)
+    {
+      Console.WriteLine(s);
+    }
+    Console.WriteLine("------------");
+
+    string[] testFiles =
+        new string[] {
+          "./sr-data/test/fangjun-test-sr-1.wav",
+          "./sr-data/test/leijun-test-sr-1.wav",
+          "./sr-data/test/liudehua-test-sr-1.wav"
+        };
+
+    float threshold = 0.6f;
+    foreach (var file in testFiles)
+    {
+      float[] embedding = ComputeEmbedding(extractor, file);
+
+      String name = manager.Search(embedding, threshold);
+      if (name == "")
+      {
+        name = "<Unknown>";
+      }
+      Console.WriteLine("{0}: {1}", file, name);
+    }
+
+    // test verify
+    if (!manager.Verify("fangjun", ComputeEmbedding(extractor, testFiles[0]), threshold))
+    {
+      Console.WriteLine("testFiles[0] should match fangjun!");
+      return;
+    }
+
+    if (!manager.Remove("fangjun"))
+    {
+      Console.WriteLine("Failed to remove fangjun");
+      return;
+    }
+
+    if (manager.Verify("fangjun", ComputeEmbedding(extractor, testFiles[0]), threshold))
+    {
+      Console.WriteLine("{0} should match no one!", testFiles[0]);
+      return;
+    }
+
+    if (manager.NumSpeakers != 1)
+    {
+      Console.WriteLine("There should only 1 speaker left.");
+      return;
+    }
+  }
+}

dotnet-examples/speaker-identification/WaveReader.cs

@@ -0,0 +1 @@
+../offline-decode-files/WaveReader.cs

dotnet-examples/speaker-identification/run.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [ ! -e ./3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-recongition-models/3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx
+fi
+
+if [ ! -d ./sr-data ]; then
+  git clone https://github.com/csukuangfj/sr-data
+fi
+
+dotnet run

dotnet-examples/speaker-identification/speaker-identification.csproj

@@ -0,0 +1,15 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <OutputType>Exe</OutputType>
+    <TargetFramework>net6.0</TargetFramework>
+    <RootNamespace>speaker_identification</RootNamespace>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <Nullable>enable</Nullable>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="org.k2fsa.sherpa.onnx" Version="*" />
+  </ItemGroup>
+
+</Project>

scripts/dotnet/examples/speaker-identification.csproj

@@ -0,0 +1,19 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <OutputType>Exe</OutputType>
+    <TargetFramework>net6.0</TargetFramework>
+    <RootNamespace>speaker_identification</RootNamespace>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <Nullable>enable</Nullable>
+  </PropertyGroup>
+
+  <PropertyGroup>
+    <RestoreSources>/tmp/packages;$(RestoreSources);https://api.nuget.org/v3/index.json</RestoreSources>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="org.k2fsa.sherpa.onnx" Version="*" />
+  </ItemGroup>
+
+</Project>