k2-fsa · csukuangfj · Jul 10, 2024 · Jul 10, 2024 · Jul 10, 2024 · Jul 10, 2024
diff --git a/.github/scripts/test-dot-net.sh b/.github/scripts/test-dot-net.sh
@@ -2,7 +2,10 @@
 
 cd dotnet-examples/
 
-cd ./online-decode-files
+cd ./keyword-spotting-from-files
+./run.sh
+
+cd ../online-decode-files
 ./run-transducer-itn.sh
 ./run-zipformer2-ctc.sh
 ./run-transducer.sh

diff --git a/.github/scripts/test-offline-transducer.sh b/.github/scripts/test-offline-transducer.sh
@@ -139,7 +139,7 @@ time $EXE \
 time $EXE \
   --tokens=$repo/tokens.txt \
   --encoder=$repo/encoder-epoch-99-avg-1.int8.onnx \
-  --decoder=$repo/decoder-epoch-99-avg-1.int8.onnx \
+  --decoder=$repo/decoder-epoch-99-avg-1.onnx \
   --joiner=$repo/joiner-epoch-99-avg-1.int8.onnx \
   --num-threads=2 \
   $repo/test_wavs/0.wav \
@@ -172,7 +172,7 @@ time $EXE \
 time $EXE \
   --tokens=$repo/tokens.txt \
   --encoder=$repo/encoder-epoch-99-avg-1.int8.onnx \
-  --decoder=$repo/decoder-epoch-99-avg-1.int8.onnx \
+  --decoder=$repo/decoder-epoch-99-avg-1.onnx \
   --joiner=$repo/joiner-epoch-99-avg-1.int8.onnx \
   --num-threads=2 \
   $repo/test_wavs/0.wav \

diff --git a/.github/scripts/test-online-transducer.sh b/.github/scripts/test-online-transducer.sh
@@ -86,7 +86,7 @@ for wave in ${waves[@]}; do
   time $EXE \
   --tokens=$repo/tokens.txt \
   --encoder=$repo/encoder-epoch-99-avg-1.int8.onnx \
-  --decoder=$repo/decoder-epoch-99-avg-1.int8.onnx \
+  --decoder=$repo/decoder-epoch-99-avg-1.onnx \
   --joiner=$repo/joiner-epoch-99-avg-1.int8.onnx \
   --num-threads=2 \
   $wave
@@ -126,7 +126,7 @@ for wave in ${waves[@]}; do
   time $EXE \
   --tokens=$repo/tokens.txt \
   --encoder=$repo/encoder-epoch-11-avg-1.int8.onnx \
-  --decoder=$repo/decoder-epoch-11-avg-1.int8.onnx \
+  --decoder=$repo/decoder-epoch-11-avg-1.onnx \
   --joiner=$repo/joiner-epoch-11-avg-1.int8.onnx \
   --num-threads=2 \
   $wave
@@ -168,7 +168,7 @@ for wave in ${waves[@]}; do
   time $EXE \
   --tokens=$repo/tokens.txt \
   --encoder=$repo/encoder-epoch-99-avg-1.int8.onnx \
-  --decoder=$repo/decoder-epoch-99-avg-1.int8.onnx \
+  --decoder=$repo/decoder-epoch-99-avg-1.onnx \
   --joiner=$repo/joiner-epoch-99-avg-1.int8.onnx \
   --num-threads=2 \
   $wave
@@ -210,7 +210,7 @@ for wave in ${waves[@]}; do
   time $EXE \
   --tokens=$repo/tokens.txt \
   --encoder=$repo/encoder-epoch-99-avg-1.int8.onnx \
-  --decoder=$repo/decoder-epoch-99-avg-1.int8.onnx \
+  --decoder=$repo/decoder-epoch-99-avg-1.onnx \
   --joiner=$repo/joiner-epoch-99-avg-1.int8.onnx \
   --num-threads=2 \
   $wave
@@ -231,7 +231,7 @@ if [ $EXE == "sherpa-onnx-ffmpeg" ]; then
   time $EXE \
   $repo/tokens.txt \
   $repo/encoder-epoch-99-avg-1.int8.onnx \
-  $repo/decoder-epoch-99-avg-1.int8.onnx \
+  $repo/decoder-epoch-99-avg-1.onnx \
   $repo/joiner-epoch-99-avg-1.int8.onnx \
   https://huggingface.co/csukuangfj/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/resolve/main/test_wavs/4.wav \
   2
@@ -271,7 +271,7 @@ for wave in ${waves[@]}; do
   time $EXE \
   --tokens=$repo/tokens.txt \
   --encoder=$repo/encoder-epoch-99-avg-1.int8.onnx \
-  --decoder=$repo/decoder-epoch-99-avg-1.int8.onnx \
+  --decoder=$repo/decoder-epoch-99-avg-1.onnx \
   --joiner=$repo/joiner-epoch-99-avg-1.int8.onnx \
   --num-threads=2 \
   $wave

diff --git a/.github/scripts/test-python.sh b/.github/scripts/test-python.sh
@@ -125,12 +125,15 @@ for name in ${wenet_models[@]}; do
   repo=$name
   log "Start testing ${repo_url}"
 
-  python3 ./python-api-examples/offline-decode-files.py \
-    --tokens=$repo/tokens.txt \
-    --wenet-ctc=$repo/model.onnx \
-    $repo/test_wavs/0.wav \
-    $repo/test_wavs/1.wav \
-    $repo/test_wavs/8k.wav
+  if false; then
+    # offline wenet ctc models are not supported by onnxruntime >= 1.18
+    python3 ./python-api-examples/offline-decode-files.py \
+      --tokens=$repo/tokens.txt \
+      --wenet-ctc=$repo/model.onnx \
+      $repo/test_wavs/0.wav \
+      $repo/test_wavs/1.wav \
+      $repo/test_wavs/8k.wav
+  fi
 
   python3 ./python-api-examples/online-decode-files.py \
     --tokens=$repo/tokens.txt \

diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,6 +1,7 @@
 ## 1.10.13
 
 * Update onnxruntime from 1.17.1 to 1.18.0
+* Add C# API for Keyword spotting
 
 ## 1.10.12
 

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -21,11 +21,6 @@ if (CMAKE_VERSION VERSION_GREATER_EQUAL "3.24.0")
   cmake_policy(SET CMP0135 NEW)
 endif()
 
-
-
-
-
-
 option(SHERPA_ONNX_ENABLE_PYTHON "Whether to build Python" OFF)
 option(SHERPA_ONNX_ENABLE_TESTS "Whether to build tests" OFF)
 option(SHERPA_ONNX_ENABLE_CHECK "Whether to build with assert" OFF)

diff --git a/dotnet-examples/keyword-spotting-from-files/Program.cs b/dotnet-examples/keyword-spotting-from-files/Program.cs
@@ -0,0 +1,99 @@
+// Copyright (c)  2024  Xiaomi Corporation
+//
+// This file shows how to do keyword spotting with sherpa-onnx.
+//
+// 1. Download a model from
+// https://github.com/k2-fsa/sherpa-onnx/releases/tag/kws-models
+//
+// wget https://github.com/k2-fsa/sherpa-onnx/releases/download/kws-models/sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01.tar.bz2
+// tar xvf sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01.tar.bz2
+//
+// 2. Now run it
+//
+// dotnet run
+
+using SherpaOnnx;
+using System.Collections.Generic;
+using System;
+
+class KeywordSpotterDemo
+{
+  static void Main(string[] args)
+  {
+    var config = new KeywordSpotterConfig();
+    config.FeatConfig.SampleRate = 16000;
+    config.FeatConfig.FeatureDim = 80;
+
+    config.ModelConfig.Transducer.Encoder = "./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01/encoder-epoch-12-avg-2-chunk-16-left-64.onnx";
+    config.ModelConfig.Transducer.Decoder = "./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01/decoder-epoch-12-avg-2-chunk-16-left-64.onnx";
+    config.ModelConfig.Transducer.Joiner = "./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01/joiner-epoch-12-avg-2-chunk-16-left-64.onnx";
+
+    config.ModelConfig.Tokens = "./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01/tokens.txt";
+    config.ModelConfig.Provider = "cpu";
+    config.ModelConfig.NumThreads = 1;
+    config.ModelConfig.Debug = 1;
+    config.KeywordsFile = "./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01/test_wavs/test_keywords.txt";
+
+    var kws = new KeywordSpotter(config);
+
+    var filename = "./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01/test_wavs/3.wav";
+
+    WaveReader waveReader = new WaveReader(filename);
+
+    Console.WriteLine("----------Use pre-defined keywords----------");
+
+    OnlineStream s = kws.CreateStream();
+    s.AcceptWaveform(waveReader.SampleRate, waveReader.Samples);
+
+    float[] tailPadding = new float[(int)(waveReader.SampleRate * 0.3)];
+    s.AcceptWaveform(waveReader.SampleRate, tailPadding);
+    s.InputFinished();
+
+    while (kws.IsReady(s))
+    {
+      kws.Decode(s);
+      var result = kws.GetResult(s);
+      if (result.Keyword != "")
+      {
+        Console.WriteLine("Detected: {0}", result.Keyword);
+      }
+    }
+
+    Console.WriteLine("----------Use pre-defined keywords + add a new keyword----------");
+    s = kws.CreateStream("y ǎn y uán @演员");
+    s.AcceptWaveform(waveReader.SampleRate, waveReader.Samples);
+
+    s.AcceptWaveform(waveReader.SampleRate, tailPadding);
+    s.InputFinished();
+
+    while (kws.IsReady(s))
+    {
+      kws.Decode(s);
+      var result = kws.GetResult(s);
+      if (result.Keyword != "")
+      {
+        Console.WriteLine("Detected: {0}", result.Keyword);
+      }
+    }
+
+    Console.WriteLine("----------Use pre-defined keywords + add 2 new keywords----------");
+
+    // Note keywords are separated by /
+    s = kws.CreateStream("y ǎn y uán @演员/zh ī m íng @知名");
+    s.AcceptWaveform(waveReader.SampleRate, waveReader.Samples);
+
+    s.AcceptWaveform(waveReader.SampleRate, tailPadding);
+    s.InputFinished();
+
+    while (kws.IsReady(s))
+    {
+      kws.Decode(s);
+      var result = kws.GetResult(s);
+      if (result.Keyword != "")
+      {
+        Console.WriteLine("Detected: {0}", result.Keyword);
+      }
+    }
+  }
+}
+
diff --git a/dotnet-examples/keyword-spotting-from-files/keyword-spotting-from-files.csproj b/dotnet-examples/keyword-spotting-from-files/keyword-spotting-from-files.csproj
@@ -0,0 +1,15 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <OutputType>Exe</OutputType>
+    <TargetFramework>net6.0</TargetFramework>
+    <RootNamespace>keyword_spotting_from_files</RootNamespace>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <Nullable>enable</Nullable>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <ProjectReference Include="..\Common\Common.csproj" />
+  </ItemGroup>
+
+</Project>
diff --git a/dotnet-examples/keyword-spotting-from-files/run.sh b/dotnet-examples/keyword-spotting-from-files/run.sh
@@ -0,0 +1,11 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [ ! -f ./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01/tokens.txt ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/kws-models/sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01.tar.bz2
+  tar xvf sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01.tar.bz2
+  rm sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01.tar.bz2
+fi
+
+dotnet run -c Release
diff --git a/dotnet-examples/online-decode-files/run-transducer-itn.sh b/dotnet-examples/online-decode-files/run-transducer-itn.sh
@@ -22,7 +22,7 @@ fi
 dotnet run -c Release \
   --tokens ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/tokens.txt \
   --encoder ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/encoder-epoch-99-avg-1.int8.onnx \
-  --decoder ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/decoder-epoch-99-avg-1.int8.onnx \
+  --decoder ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/decoder-epoch-99-avg-1.onnx \
   --joiner ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/joiner-epoch-99-avg-1.int8.onnx \
   --rule-fsts ./itn_zh_number.fst \
   --decoding-method greedy_search \

diff --git a/dotnet-examples/online-decode-files/run-transducer.sh b/dotnet-examples/online-decode-files/run-transducer.sh
@@ -14,7 +14,7 @@ fi
 dotnet run -c Release \
   --tokens ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/tokens.txt \
   --encoder ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/encoder-epoch-99-avg-1.int8.onnx \
-  --decoder ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/decoder-epoch-99-avg-1.int8.onnx \
+  --decoder ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/decoder-epoch-99-avg-1.onnx \
   --joiner ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/joiner-epoch-99-avg-1.int8.onnx \
   --decoding-method greedy_search \
   --files ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/test_wavs/1.wav \

diff --git a/dotnet-examples/sherpa-onnx.sln b/dotnet-examples/sherpa-onnx.sln
@@ -25,6 +25,8 @@ Project("{9A19103F-16F7-4668-BE54-9A1E7A4F7556}") = "vad-non-streaming-asr-paraf
 EndProject
 Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "Common", "Common\Common.csproj", "{401E963F-E25A-43CE-987D-8DB2D4715756}"
 EndProject
+Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "keyword-spotting-from-files", "keyword-spotting-from-files\keyword-spotting-from-files.csproj", "{A87EDD31-D654-4C9F-AED7-F6F2825659BD}"
+EndProject
 Global
 	GlobalSection(SolutionConfigurationPlatforms) = preSolution
 		Debug|Any CPU = Debug|Any CPU
@@ -75,6 +77,10 @@ Global
 		{401E963F-E25A-43CE-987D-8DB2D4715756}.Debug|Any CPU.Build.0 = Debug|Any CPU
 		{401E963F-E25A-43CE-987D-8DB2D4715756}.Release|Any CPU.ActiveCfg = Release|Any CPU
 		{401E963F-E25A-43CE-987D-8DB2D4715756}.Release|Any CPU.Build.0 = Release|Any CPU
+		{A87EDD31-D654-4C9F-AED7-F6F2825659BD}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{A87EDD31-D654-4C9F-AED7-F6F2825659BD}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{A87EDD31-D654-4C9F-AED7-F6F2825659BD}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{A87EDD31-D654-4C9F-AED7-F6F2825659BD}.Release|Any CPU.Build.0 = Release|Any CPU
 	EndGlobalSection
 	GlobalSection(SolutionProperties) = preSolution
 		HideSolutionNode = FALSE

diff --git a/dotnet-examples/speech-recognition-from-microphone/run-transducer.sh b/dotnet-examples/speech-recognition-from-microphone/run-transducer.sh
@@ -18,5 +18,5 @@ fi
 dotnet run -c Release \
   --tokens ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/tokens.txt \
   --encoder ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/encoder-epoch-99-avg-1.int8.onnx \
-  --decoder ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/decoder-epoch-99-avg-1.int8.onnx \
+  --decoder ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/decoder-epoch-99-avg-1.onnx \
   --joiner ./sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/joiner-epoch-99-avg-1.int8.onnx
diff --git a/flutter/sherpa_onnx/example/example.md b/flutter/sherpa_onnx/example/example.md
@@ -5,6 +5,7 @@
 | Functions | URL | Supported Platforms|
 |---|---|---|
 |Streaming speech recognition| [Address](https://github.com/k2-fsa/sherpa-onnx/tree/master/flutter-examples/streaming_asr)| Android, macOS, Windows|
+|Speech synthesis| [Address](https://github.com/k2-fsa/sherpa-onnx/tree/master/flutter-examples/tts)| Android, iOS, Linux, macOS, Windows|
 
 ## Pure dart-examples
 

diff --git a/scripts/dotnet/.gitignore b/scripts/dotnet/.gitignore
@@ -1,8 +1,10 @@
 all
 macos-arm64
 macos-x64
-linux
-windows
+linux-x64
+linux-arm64
+windows-arm64
 windows-x64
 windows-x86
 packages
+tmp
diff --git a/scripts/dotnet/KeywordResult.cs b/scripts/dotnet/KeywordResult.cs
@@ -0,0 +1,44 @@
+/// Copyright (c)  2024  Xiaomi Corporation
+
+using System;
+using System.Runtime.InteropServices;
+using System.Text;
+
+namespace SherpaOnnx
+{
+    public class KeywordResult
+    {
+        public KeywordResult(IntPtr handle)
+        {
+            Impl impl = (Impl)Marshal.PtrToStructure(handle, typeof(Impl));
+
+            // PtrToStringUTF8() requires .net standard 2.1
+            // _keyword = Marshal.PtrToStringUTF8(impl.Keyword);
+
+            int length = 0;
+
+            unsafe
+            {
+                byte* buffer = (byte*)impl.Keyword;
+                while (*buffer != 0)
+                {
+                    ++buffer;
+                    length += 1;
+                }
+            }
+
+            byte[] stringBuffer = new byte[length];
+            Marshal.Copy(impl.Keyword, stringBuffer, 0, length);
+            _keyword = Encoding.UTF8.GetString(stringBuffer);
+        }
+
+        [StructLayout(LayoutKind.Sequential)]
+        struct Impl
+        {
+            public IntPtr Keyword;
+        }
+
+        private String _keyword;
+        public String Keyword => _keyword;
+    }
+}