Add VAD demo for Java API (#928)

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/SileroVadModelConfig.java

@@ -0,0 +1,81 @@
+// Copyright 2024 Xiaomi Corporation
+
+package com.k2fsa.sherpa.onnx;
+
+public class SileroVadModelConfig {
+    private final String model;
+    private final float threshold;
+    private final float minSilenceDuration;
+    private final float minSpeechDuration;
+    private final int windowSize;
+
+    private SileroVadModelConfig(Builder builder) {
+        this.model = builder.model;
+        this.threshold = builder.threshold;
+        this.minSilenceDuration = builder.minSilenceDuration;
+        this.minSpeechDuration = builder.minSpeechDuration;
+        this.windowSize = builder.windowSize;
+    }
+
+    public static Builder builder() {
+        return new Builder();
+    }
+
+    public String getModel() {
+        return model;
+    }
+
+    public float getThreshold() {
+        return threshold;
+    }
+
+    public float getMinSilenceDuration() {
+        return minSilenceDuration;
+    }
+
+    public float getMinSpeechDuration() {
+        return minSpeechDuration;
+    }
+
+    public int getWindowSize() {
+        return windowSize;
+    }
+
+    public static class Builder {
+        private String model = "";
+        private float threshold = 0.5f;
+        private float minSilenceDuration = 0.25f;
+        private float minSpeechDuration = 0.5f;
+        private int windowSize = 512;
+
+        public SileroVadModelConfig build() {
+            return new SileroVadModelConfig(this);
+        }
+
+
+        public Builder setModel(String model) {
+            this.model = model;
+            return this;
+        }
+
+        public Builder setThreshold(float threshold) {
+            this.threshold = threshold;
+            return this;
+        }
+
+        public Builder setMinSilenceDuration(float minSilenceDuration) {
+            this.minSilenceDuration = minSilenceDuration;
+            return this;
+        }
+
+        public Builder setMinSpeechDuration(float minSpeechDuration) {
+            this.minSpeechDuration = minSpeechDuration;
+            return this;
+        }
+
+        public Builder setWindowSize(int windowSize) {
+            this.windowSize = windowSize;
+            return this;
+        }
+    }
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/SpeechSegment.java

@@ -0,0 +1,20 @@
+package com.k2fsa.sherpa.onnx;
+
+public class SpeechSegment {
+
+    private final int start;
+    private final float[] samples;
+
+    public SpeechSegment(int start, float[] samples) {
+        this.start = start;
+        this.samples = samples;
+    }
+
+    public int getStart() {
+        return start;
+    }
+
+    public float[] getSamples() {
+        return samples;
+    }
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/Vad.java

@@ -0,0 +1,78 @@
+// Copyright 2024 Xiaomi Corporation
+
+package com.k2fsa.sherpa.onnx;
+
+public class Vad {
+    static {
+        System.loadLibrary("sherpa-onnx-jni");
+    }
+
+    private long ptr = 0;
+
+    public Vad(VadModelConfig config) {
+        ptr = newFromFile(config);
+    }
+
+    @Override
+    protected void finalize() throws Throwable {
+        release();
+    }
+
+    public void release() {
+        if (this.ptr == 0) {
+            return;
+        }
+        delete(this.ptr);
+        this.ptr = 0;
+    }
+
+    public void acceptWaveform(float[] samples) {
+        acceptWaveform(this.ptr, samples);
+    }
+
+    public boolean empty() {
+        return empty(this.ptr);
+    }
+
+    public void pop() {
+        pop(this.ptr);
+    }
+
+    public void clear() {
+        clear(this.ptr);
+    }
+
+    public void reset() {
+        reset(this.ptr);
+    }
+
+    public SpeechSegment front() {
+        Object[] arr = front(this.ptr);
+        int start = (int) arr[0];
+        float[] samples = (float[]) arr[1];
+
+        return new SpeechSegment(start, samples);
+    }
+
+    public boolean isSpeechDetected() {
+        return isSpeechDetected(this.ptr);
+    }
+
+    private native void delete(long ptr);
+
+    private native long newFromFile(VadModelConfig config);
+
+    private native void acceptWaveform(long ptr, float[] samples);
+
+    private native boolean empty(long ptr);
+
+    private native void pop(long ptr);
+
+    private native void clear(long ptr);
+
+    private native Object[] front(long ptr);
+
+    private native boolean isSpeechDetected(long ptr);
+
+    private native void reset(long ptr);
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/VadModelConfig.java

@@ -0,0 +1,80 @@
+// Copyright 2024 Xiaomi Corporation
+
+package com.k2fsa.sherpa.onnx;
+
+public class VadModelConfig {
+    private final SileroVadModelConfig sileroVadModelConfig;
+    private final int sampleRate;
+    private final int numThreads;
+    private final boolean debug;
+    private final String provider;
+
+    private VadModelConfig(Builder builder) {
+        this.sileroVadModelConfig = builder.sileroVadModelConfig;
+        this.sampleRate = builder.sampleRate;
+        this.numThreads = builder.numThreads;
+        this.debug = builder.debug;
+        this.provider = builder.provider;
+    }
+
+    public static Builder builder() {
+        return new Builder();
+    }
+
+    public SileroVadModelConfig getSileroVadModelConfig() {
+        return sileroVadModelConfig;
+    }
+
+    public int getSampleRate() {
+        return sampleRate;
+    }
+
+    public int getNumThreads() {
+        return numThreads;
+    }
+
+    public String getProvider() {
+        return provider;
+    }
+
+    public boolean getDebug() {
+        return debug;
+    }
+
+    public static class Builder {
+        private SileroVadModelConfig sileroVadModelConfig = new SileroVadModelConfig.Builder().build();
+        private int sampleRate = 16000;
+        private int numThreads = 1;
+        private boolean debug = true;
+        private String provider = "cpu";
+
+        public VadModelConfig build() {
+            return new VadModelConfig(this);
+        }
+
+        public Builder setSileroVadModelConfig(SileroVadModelConfig sileroVadModelConfig) {
+            this.sileroVadModelConfig = sileroVadModelConfig;
+            return this;
+        }
+
+        public Builder setSampleRate(int sampleRate) {
+            this.sampleRate = sampleRate;
+            return this;
+        }
+
+        public Builder setNumThreads(int numThreads) {
+            this.numThreads = numThreads;
+            return this;
+        }
+
+        public Builder setDebug(boolean debug) {
+            this.debug = debug;
+            return this;
+        }
+
+        public Builder setProvider(String provider) {
+            this.provider = provider;
+            return this;
+        }
+    }
+}

sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/WaveWriter.java

@@ -0,0 +1,15 @@
+// Copyright 2024 Xiaomi Corporation
+
+package com.k2fsa.sherpa.onnx;
+
+public class WaveWriter {
+    public WaveWriter() {
+    }
+
+    public static boolean write(String filename, float[] samples, int sampleRate) {
+        WaveWriter w = new WaveWriter();
+        return w.writeWaveToFile(filename, samples, sampleRate);
+    }
+
+    private native boolean writeWaveToFile(String filename, float[] samples, int sampleRate);
+}