java使用vosk开源框架完成离线版语音识别，中文语言模型可训练

vosk 官网

1. 链接： https://alphacephei.com/vosk/

2. 可以下载中文模型
   

3. 有训练的方法
   

java使用vosk

1. 引入依赖

    <!-- 获取音频信息 -->
    <dependency>
      <groupId>org</groupId>
      <artifactId>jaudiotagger</artifactId>
      <version>2.0.3</version>
    </dependency>

<!-- 语音识别 -->
<dependency>
    <groupId>net.java.dev.jna</groupId>
    <artifactId>jna</artifactId>
    <version>5.7.0</version>
</dependency>
<dependency>
    <groupId>com.alphacephei</groupId>
    <artifactId>vosk</artifactId>
    <version>0.3.32</version>
</dependency>

2. 使用方法

    public static void main(String[] args) {
        LibVosk.setLogLevel(LogLevel.DEBUG);
        File file = new File("D:\\test.wav");//语音文件仅支持wav
        try (Model model = new Model(System.getProperty("user.dir")+ "/src/main/resources/vosk-model-small-cn-0.22");//模型地址
             InputStream ais = AudioSystem.getAudioInputStream(new BufferedInputStream(new FileInputStream(file))); 
             Recognizer recognizer = new Recognizer(model, getSampleRate(file))) { 

            //recognizer.setWords(true); // 开启词信息的详细输出

            int bytes;
            byte[] b = new byte[4096];//缓存大小：4k
            while ((bytes = ais.read(b)) >= 0) {
                recognizer.acceptWaveForm(b, bytes);
            }
            System.out.println(recognizer.getFinalResult() );
        } catch (Exception e) {
            e.printStackTrace();
        }
    }

    /**
     * 获取采样率，采样率为音频采样率的声道倍数
     */
    private static  Float getSampleRate(File file) throws Exception {
        WavFileReader fileReader = new WavFileReader();
        AudioFile audioFile = fileReader.read(file);
        String sampleRate = audioFile.getAudioHeader().getSampleRate();
        String channel = audioFile.getAudioHeader().getChannels();
        System.out.println("语音的声道：" + channel );
        System.out.println("语音的采样率：" + sampleRate);
        return Float.parseFloat(sampleRate) * Integer.parseInt(channel);
    }