Add Java API for spoken language identification with whisper multilingual models (#817)

java-api-examples/.gitignore

@@ -1,4 +1,3 @@
 lib
 hs_err*
-!run-streaming*.sh
-!run-non-streaming*.sh
+!run-*.sh

java-api-examples/README.md

@@ -29,3 +29,9 @@ This directory contains examples for the JAVA API of sherpa-onnx.
 ./run-non-streaming-tts-coqui-de.sh
 ./run-non-streaming-tts-vits-zh.sh
 ```
+
+## Spoken language identification
+
+```bash
+./run-spoken-language-identification-whisper.sh
+```

java-api-examples/SpokenLanguageIdentificationWhisper.java

@@ -0,0 +1,61 @@
+// Copyright 2024 Xiaomi Corporation
+
+// This file shows how to use a multilingual whisper model for
+// spoken language identification.
+//
+// Note that it needs a multilingual whisper model. For instance,
+// tiny works, but tiny.en doesn't.
+import com.k2fsa.sherpa.onnx.*;
+
+public class SpokenLanguageIdentificationWhisper {
+  public static void main(String[] args) {
+    // please download model and test files from
+    // https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models
+    String encoder = "./sherpa-onnx-whisper-tiny/tiny-encoder.int8.onnx";
+    String decoder = "./sherpa-onnx-whisper-tiny/tiny-decoder.int8.onnx";
+
+    String[] testFiles =
+        new String[] {
+          "./spoken-language-identification-test-wavs/en-english.wav",
+          "./spoken-language-identification-test-wavs/de-german.wav",
+          "./spoken-language-identification-test-wavs/zh-chinese.wav",
+          "./spoken-language-identification-test-wavs/es-spanish.wav",
+          "./spoken-language-identification-test-wavs/fa-persian.wav",
+          "./spoken-language-identification-test-wavs/ko-korean.wav",
+          "./spoken-language-identification-test-wavs/ja-japanese.wav",
+          "./spoken-language-identification-test-wavs/ru-russian.wav",
+          "./spoken-language-identification-test-wavs/uk-ukrainian.wav",
+        };
+
+    SpokenLanguageIdentificationWhisperConfig whisper =
+        SpokenLanguageIdentificationWhisperConfig.builder()
+            .setEncoder(encoder)
+            .setDecoder(decoder)
+            .build();
+
+    SpokenLanguageIdentificationConfig config =
+        SpokenLanguageIdentificationConfig.builder()
+            .setWhisper(whisper)
+            .setNumThreads(1)
+            .setDebug(true)
+            .build();
+
+    SpokenLanguageIdentification slid = new SpokenLanguageIdentification(config);
+    for (String filename : testFiles) {
+      WaveReader reader = new WaveReader(filename);
+
+      OfflineStream stream = slid.createStream();
+      stream.acceptWaveform(reader.getSamples(), reader.getSampleRate());
+
+      String lang = slid.compute(stream);
+      System.out.println("---");
+      System.out.printf("filename: %s\n", filename);
+      System.out.printf("lang: %s\n", lang);
+
+      stream.release();
+    }
+    System.out.println("---");
+
+    slid.release();
+  }
+}

java-api-examples/run-spoken-language-identification-whisper.sh

@@ -0,0 +1,59 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [[ ! -f ../build/lib/libsherpa-onnx-jni.dylib  && ! -f ../build/lib/libsherpa-onnx-jni.so ]]; then
+  mkdir -p ../build
+  pushd ../build
+  cmake \
+    -DSHERPA_ONNX_ENABLE_PYTHON=OFF \
+    -DSHERPA_ONNX_ENABLE_TESTS=OFF \
+    -DSHERPA_ONNX_ENABLE_CHECK=OFF \
+    -DBUILD_SHARED_LIBS=ON \
+    -DSHERPA_ONNX_ENABLE_PORTAUDIO=OFF \
+    -DSHERPA_ONNX_ENABLE_JNI=ON \
+    ..
+
+  make -j4
+  ls -lh lib
+  popd
+fi
+
+if [ ! -f ../sherpa-onnx/java-api/build/sherpa-onnx.jar ]; then
+  pushd ../sherpa-onnx/java-api
+  make
+  popd
+fi
+
+if [[ ! -f ../build/lib/libsherpa-onnx-jni.dylib  && ! -f ../build/lib/libsherpa-onnx-jni.so ]]; then
+  cmake \
+    -DSHERPA_ONNX_ENABLE_PYTHON=OFF \
+    -DSHERPA_ONNX_ENABLE_TESTS=OFF \
+    -DSHERPA_ONNX_ENABLE_CHECK=OFF \
+    -DBUILD_SHARED_LIBS=ON \
+    -DSHERPA_ONNX_ENABLE_PORTAUDIO=OFF \
+    -DSHERPA_ONNX_ENABLE_JNI=ON \
+    ..
+
+  make -j4
+  ls -lh lib
+fi
+
+# Note that it needs a multilingual whisper model. so, for example, tiny works while tiny.en does not work
+# https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-whisper-tiny.tar.bz2
+if [ ! -f ./sherpa-onnx-whisper-tiny/tiny-encoder.int8.onnx ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-whisper-tiny.tar.bz2
+  tar xvf sherpa-onnx-whisper-tiny.tar.bz2
+  rm sherpa-onnx-whisper-tiny.tar.bz2
+fi
+
+if [ ! -f ./spoken-language-identification-test-wavs/en-english.wav ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/spoken-language-identification-test-wavs.tar.bz2
+  tar xvf spoken-language-identification-test-wavs.tar.bz2
+  rm spoken-language-identification-test-wavs.tar.bz2
+fi
+
+java \
+  -Djava.library.path=$PWD/../build/lib \
+  -cp ../sherpa-onnx/java-api/build/sherpa-onnx.jar \
+  ./SpokenLanguageIdentificationWhisper.java