EngineX-Iluvatar/enginex_bi_series-sherpa-onnx
Archived
15
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

C# API for speaker diarization (#1407)

Browse Source
This commit is contained in:
Fangjun Kuang
2024-10-10 14:29:05 +08:00
committed by GitHub
parent bd50e79590
commit a45e5dba99
12 changed files with 408 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

83
dotnet-examples/offline-speaker-diarization/Program.cs Normal file
View File

@@ -0,0 +1,83 @@
// Copyright (c) 2024 Xiaomi Corporation
//
// This file shows how to use sherpa-onnx C# API for speaker diarization
/*
Usage:
Step 1: Download a speaker segmentation model
Please visit https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-segmentation-models
for a list of available models. The following is an example
wget https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-segmentation-models/sherpa-onnx-pyannote-segmentation-3-0.tar.bz2
tar xvf sherpa-onnx-pyannote-segmentation-3-0.tar.bz2
rm sherpa-onnx-pyannote-segmentation-3-0.tar.bz2
Step 2: Download a speaker embedding extractor model
Please visit https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-recongition-models
for a list of available models. The following is an example
wget https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-recongition-models/3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx
Step 3. Download test wave files
Please visit https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-segmentation-models
for a list of available test wave files. The following is an example
wget https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-segmentation-models/0-four-speakers-zh.wav
Step 4. Run it
dotnet run
*/
using SherpaOnnx;
using System;
class OfflineSpeakerDiarizationDemo
{
static void Main(string[] args)
{
var config = new OfflineSpeakerDiarizationConfig();
config.Segmentation.Pyannote.Model = "./sherpa-onnx-pyannote-segmentation-3-0/model.onnx";
config.Embedding.Model = "./3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx";
// the test wave ./0-four-speakers-zh.wav has 4 speakers, so
// we set num_clusters to 4
//
config.Clustering.NumClusters = 4;
// If you don't know the number of speakers in the test wave file, please
// use
// config.Clustering.Threshold = 0.5; // You need to tune this threshold
var sd = new OfflineSpeakerDiarization(config);
var testWaveFile = "./0-four-speakers-zh.wav";
WaveReader waveReader = new WaveReader(testWaveFile);
if (sd.SampleRate != waveReader.SampleRate)
{
Console.WriteLine($"Expected sample rate: {sd.SampleRate}. Given: {waveReader.SampleRate}");
return;
}
Console.WriteLine("Started");
// var segments = sd.Process(waveReader.Samples); // this one is also ok
var MyProgressCallback = (int numProcessedChunks, int numTotalChunks, IntPtr arg) =>
{
float progress = 100.0F * numProcessedChunks / numTotalChunks;
Console.WriteLine("Progress {0}%", String.Format("{0:0.00}", progress));
return 0;
};
var callback = new OfflineSpeakerDiarizationProgressCallback(MyProgressCallback);
var segments = sd.ProcessWithCallback(waveReader.Samples, callback, IntPtr.Zero);
foreach (var s in segments)
{
Console.WriteLine("{0} -- {1} speaker_{2}", String.Format("{0:0.00}", s.Start), String.Format("{0:0.00}", s.End), s.Speaker);
}
}
}

15
dotnet-examples/offline-speaker-diarization/offline-speaker-diarization.csproj Normal file
View File

@@ -0,0 +1,15 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
<TargetFramework>net6.0</TargetFramework>
<RootNamespace>offline_speaker_diarization</RootNamespace>
<ImplicitUsings>enable</ImplicitUsings>
<Nullable>enable</Nullable>
</PropertyGroup>
<ItemGroup>
<ProjectReference Include="..\Common\Common.csproj" />
</ItemGroup>
</Project>

18
dotnet-examples/offline-speaker-diarization/run.sh Executable file
View File

@@ -0,0 +1,18 @@
#!/usr/bin/env bash
if [ ! -f ./sherpa-onnx-pyannote-segmentation-3-0/model.onnx ]; then
curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-segmentation-models/sherpa-onnx-pyannote-segmentation-3-0.tar.bz2
tar xvf sherpa-onnx-pyannote-segmentation-3-0.tar.bz2
rm sherpa-onnx-pyannote-segmentation-3-0.tar.bz2
fi
if [ ! -f ./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 [ ! -f ./0-four-speakers-zh.wav ]; then
curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-segmentation-models/0-four-speakers-zh.wav
fi
dotnet run

6
dotnet-examples/sherpa-onnx.sln
View File

@@ -31,6 +31,8 @@ Project("{9A19103F-16F7-4668-BE54-9A1E7A4F7556}") = "keyword-spotting-from-micro
EndProject
Project("{9A19103F-16F7-4668-BE54-9A1E7A4F7556}") = "TTS", "TTS\TTS.csproj", "{DACE4A18-4FC8-4437-92BF-5A90BA81286C}"
EndProject
Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "offline-speaker-diarization", "offline-speaker-diarization\offline-speaker-diarization.csproj", "{D3A1FF28-A77D-429D-AEAC-2BA77CA682BC}"
EndProject
Global
GlobalSection(SolutionConfigurationPlatforms) = preSolution
Debug|Any CPU = Debug|Any CPU
@@ -93,6 +95,10 @@ Global
{DACE4A18-4FC8-4437-92BF-5A90BA81286C}.Debug|Any CPU.Build.0 = Debug|Any CPU
{DACE4A18-4FC8-4437-92BF-5A90BA81286C}.Release|Any CPU.ActiveCfg = Release|Any CPU
{DACE4A18-4FC8-4437-92BF-5A90BA81286C}.Release|Any CPU.Build.0 = Release|Any CPU
{D3A1FF28-A77D-429D-AEAC-2BA77CA682BC}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
{D3A1FF28-A77D-429D-AEAC-2BA77CA682BC}.Debug|Any CPU.Build.0 = Debug|Any CPU
{D3A1FF28-A77D-429D-AEAC-2BA77CA682BC}.Release|Any CPU.ActiveCfg = Release|Any CPU
{D3A1FF28-A77D-429D-AEAC-2BA77CA682BC}.Release|Any CPU.Build.0 = Release|Any CPU
EndGlobalSection
GlobalSection(SolutionProperties) = preSolution
HideSolutionNode = FALSE