摘要：項(xiàng)目中需要用到對(duì)人聲音頻數(shù)據(jù)進(jìn)行變聲處理。具體工具類(lèi)代碼變聲原始數(shù)據(jù)輸入流變速率大于為加快語(yǔ)速，小于為放慢語(yǔ)速音調(diào)變化率大于為降低音調(diào)深沉，小于為提升音調(diào)尖銳變聲后的數(shù)據(jù)輸入流這里根據(jù)自己格式修改對(duì)應(yīng)參數(shù)。

項(xiàng)目中需要用到對(duì)PCM人聲音頻數(shù)據(jù)進(jìn)行變聲處理。苦苦掙扎了一周終于找到了純Java實(shí)現(xiàn)的一套框架——TarsosDSP。功能非常強(qiáng)大！可以實(shí)時(shí)音頻處理！當(dāng)然我只用到了對(duì)文件處理。實(shí)際上邏輯是一樣的

TarsosDSP的GitHub地址：https://github.com/JorenSix/T...。將它整合至自己的項(xiàng)目工程。

具體Java工具類(lèi)代碼：

    /**
     * 變聲
     * @param rawPcmInputStream 原始PCM數(shù)據(jù)輸入流
     * @param speedFactor 變速率 (0,2) 大于1為加快語(yǔ)速，小于1為放慢語(yǔ)速
     * @param rateFactor 音調(diào)變化率 (0,2) 大于1為降低音調(diào)（深沉），小于1為提升音調(diào)（尖銳）
     * @return 變聲后的PCM數(shù)據(jù)輸入流
     */
    public static InputStream speechPitchShift(final InputStream rawPcmInputStream,double speedFactor,double rateFactor) {
        // 這里根據(jù)自己PCM格式修改對(duì)應(yīng)參數(shù)。我們項(xiàng)目里音頻格式是固定的，所以寫(xiě)死
        TarsosDSPAudioFormat format = new TarsosDSPAudioFormat(16000,16,1,true,false);
        AudioInputStream inputStream = new AudioInputStream(rawPcmInputStream, JVMAudioInputStream.toAudioFormat(format),AudioSystem.NOT_SPECIFIED);
        JVMAudioInputStream stream = new JVMAudioInputStream(inputStream);

        WaveformSimilarityBasedOverlapAdd w = new WaveformSimilarityBasedOverlapAdd(WaveformSimilarityBasedOverlapAdd.Parameters.speechDefaults(speedFactor, 16000));
        int inputBufferSize = w.getInputBufferSize();
        int overlap = w.getOverlap();
        AudioDispatcher dispatcher = new AudioDispatcher(stream, inputBufferSize ,overlap);
        w.setDispatcher(dispatcher);

        AudioOutputToByteArray out = new AudioOutputToByteArray();

        dispatcher.addAudioProcessor(w);
        dispatcher.addAudioProcessor(new RateTransposer(rateFactor));
        dispatcher.addAudioProcessor(out);
        dispatcher.run();

        return new ByteArrayInputStream(out.getData());
    }

其中數(shù)據(jù)轉(zhuǎn)錄器(AudioOutputToByteArray)代碼如下：

public class AudioOutputToByteArray implements AudioProcessor {
    private boolean isDone = false;
    private byte[] out = null;
    private ByteArrayOutputStream bos;

    public AudioOutputToByteArray() {
        bos = new ByteArrayOutputStream();
    }

    public byte[] getData() {
        while (!isDone && out == null) {
            try {
                Thread.sleep(10);
            } catch (InterruptedException ignored) {}
        }

        return out;
    }

    @Override
    public boolean process(AudioEvent audioEvent) {
        bos.write(audioEvent.getByteBuffer(),0,audioEvent.getByteBuffer().length);
        return true;
    }

    @Override
    public void processingFinished() {
        out = bos.toByteArray().clone();
        bos = null;
        isDone = true;
    }
}

可以通過(guò)這個(gè)工具方法播放音頻：

    /**
     * 播放PCM
     *
     * 不要在非桌面環(huán)境調(diào)用。。。鬼知道會(huì)發(fā)生什么
     * @param rawPcmInputStream 原始PCM數(shù)據(jù)輸入流
     * @throws LineUnavailableException
     */
    public static void play(final InputStream rawPcmInputStream) throws LineUnavailableException {
        // 這里根據(jù)自己PCM格式修改對(duì)應(yīng)參數(shù)。我們項(xiàng)目里音頻格式是固定的，所以寫(xiě)死
        TarsosDSPAudioFormat format = new TarsosDSPAudioFormat(16000,16,1,true,false);
        AudioInputStream inputStream = new AudioInputStream(rawPcmInputStream, JVMAudioInputStream.toAudioFormat(format),AudioSystem.NOT_SPECIFIED);
        JVMAudioInputStream stream = new JVMAudioInputStream(inputStream);
        AudioDispatcher dispatcher = new AudioDispatcher(stream, 1024 ,0);
        dispatcher.addAudioProcessor(new AudioPlayer(format,1024));
        dispatcher.run();
    }