如何在Java中播放Opus编码的音频？

Question

在播放解码后的音频时，我已经成功地制作出了从咕噜声到尖叫声再到恶魔歌唱等各种各样的声音。其中最接近的声音听起来像是快进播放，而且播放时间只有约15秒。我尝试了许多解码和AudioSystem API方法的组合，但似乎没有什么效果。

那么，是什么导致了这种音频失真？

该文件的Opusinfo显示如下：

Processing file "test.opus"...

New logical stream (#1, serial: 00002c88): type opus
Encoded with libopus 1.1
User comments section follows...
     ENCODER=opusenc from opus-tools 0.1.9
Opus stream 1:
    Pre-skip: 356
    Playback gain: 0 dB
    Channels: 1
    Original sample rate: 44100Hz
    Packet duration:   20.0ms (max),   20.0ms (avg),   20.0ms (min)
    Page duration:   1000.0ms (max),  996.8ms (avg),  200.0ms (min)
    Total data length: 1930655 bytes (overhead: 1.04%)
    Playback length: 4m:09.173s
    Average bitrate: 61.99 kb/s, w/o overhead: 61.34 kb/s
Logical stream 1 ended

这个文件在使用VLC播放时可以正确播放。

我正在尝试使用以下库来解码文件：

• VorbisJava (https://github.com/Gagravarr/VorbisJava/) - 用于从OGG容器中提取帧

• LibJitsi (https://jitsi.org/Projects/LibJitsi) - 其中包含一个JNI包装器，用于解码Opus帧

下面是SSCCE

package me.justinb.mediapad.audio;

import org.gagravarr.ogg.OggFile;
import org.gagravarr.ogg.OggPacket;
import org.jitsi.impl.neomedia.codec.audio.opus.Opus;
import javax.sound.sampled.*;
import java.io.*;
import java.nio.ByteBuffer;

public class OpusAudioPlayer {
    private static int BUFFER_SIZE = 1024 * 1024;
    private static int INPUT_BITRATE = 48000;
    private static int OUTPUT_BITRATE = 44100;
    private OggFile oggFile;
    private long opusState;
    private ByteBuffer decodeBuffer = ByteBuffer.allocate(BUFFER_SIZE); 
    private AudioFormat audioFormat = new AudioFormat(OUTPUT_BITRATE, 16, 1, true, false);

    public static void main(String[] args) {
        try {
            OpusAudioPlayer opusAudioPlayer = new OpusAudioPlayer(new File("test.opus"));
            opusAudioPlayer.play();
        } catch (IOException e) {
            e.printStackTrace();
        }
    }

    public OpusAudioPlayer(File audioFile) throws IOException {
        oggFile = new OggFile(new FileInputStream(audioFile));
        opusState = Opus.decoder_create(INPUT_BITRATE, 1);
        System.out.println("Audio format: " + audioFormat);
    }

    private byte[] decode(byte[] packetData) {
        int frameSize = Opus.decoder_get_nb_samples(opusState, packetData, 0, packetData.length);
        int decodedSamples = Opus.decode(opusState, packetData, 0, packetData.length, decodeBuffer.array(), 0, frameSize, 0);
        if (decodedSamples < 0) {
            System.out.println("Decode error: " + decodedSamples);
            decodeBuffer.clear();
            return null;
        }
        decodeBuffer.position(decodedSamples * 2); // 2 bytes per sample
        decodeBuffer.flip();

        byte[] decodedData = new byte[decodeBuffer.remaining()];
        decodeBuffer.get(decodedData);
        decodeBuffer.flip();
        System.out.println(String.format("Encoded frame size: %d bytes", packetData.length));
        System.out.println(String.format("Decoded frame size: %d bytes", decodedData.length));
        System.out.println(String.format("Decoded %d samples", decodedSamples));
        return decodedData;
    }

    public void play() {
        int totalDecodedBytes = 0;
        try {
            SourceDataLine speaker = AudioSystem.getSourceDataLine(audioFormat);
            OggPacket nextPacket = oggFile.getPacketReader().getNextPacket();
            // Move to beginning of stream
            while ( !nextPacket.isBeginningOfStream()) {
                nextPacket = oggFile.getPacketReader().getNextPacket();
            }
            speaker.open();
            speaker.start();
            while(nextPacket != null) {
                // Decode each packet
                byte[] decodedData = decode(nextPacket.getData());
                if(decodedData != null) {
                    // Write packet to SourceDataLine
                    speaker.write(decodedData, 0, decodedData.length);
                    totalDecodedBytes += decodedData.length;
                }
                nextPacket = oggFile.getPacketReader().getNextPacket();
            }
            speaker.drain();
            speaker.close();
            System.out.println(String.format("Decoded to %d bytes", totalDecodedBytes));
        } catch (Exception e) {
            e.printStackTrace();
        }
    }
}

- Justin

Answer 1

我的特定问题似乎是由于VorbisJava中的一个漏洞引起的。我现在正在使用J-Ogg，它可以处理容器解析而不会出现任何问题。我确信有人会觉得这很有用。

以下是最终代码，展示如何在Java中播放Opus编码的音频：

package me.justinb.mediapad.audio;

import de.jarnbjo.ogg.FileStream;
import de.jarnbjo.ogg.LogicalOggStream;
import org.jitsi.impl.neomedia.codec.audio.opus.Opus;

import javax.sound.sampled.AudioFormat;
import javax.sound.sampled.AudioSystem;
import javax.sound.sampled.SourceDataLine;
import java.io.File;
import java.io.IOException;
import java.io.RandomAccessFile;
import java.nio.ByteBuffer;
import java.util.Collection;

public class OpusAudioPlayer {
    private static int BUFFER_SIZE = 1024 * 1024;
    private static int INPUT_BITRATE = 48000;
    private static int OUTPUT_BITRATE = 48000;

    private FileStream oggFile;
    private long opusState;

    private ByteBuffer decodeBuffer = ByteBuffer.allocate(BUFFER_SIZE);

    private AudioFormat audioFormat = new AudioFormat(OUTPUT_BITRATE, 16, 1, true, false);

    public static void main(String[] args) {
        try {
            OpusAudioPlayer opusAudioPlayer = new OpusAudioPlayer(new File("test.opus"));
            opusAudioPlayer.play();
        } catch (IOException e) {
            e.printStackTrace();
        }
    }

    public OpusAudioPlayer(File audioFile) throws IOException {
        oggFile = new FileStream(new RandomAccessFile(audioFile, "r"));
        opusState = Opus.decoder_create(INPUT_BITRATE, 1);
    }

    private byte[] decode(byte[] packetData) {
        int frameSize = Opus.decoder_get_nb_samples(opusState, packetData, 0, packetData.length);
        int decodedSamples = Opus.decode(opusState, packetData, 0, packetData.length, decodeBuffer.array(), 0, frameSize, 0);
        if (decodedSamples < 0) {
            System.out.println("Decode error: " + decodedSamples);
            decodeBuffer.clear();
            return null;
        }
        decodeBuffer.position(decodedSamples * 2); // 2 bytes per sample
        decodeBuffer.flip();

        byte[] decodedData = new byte[decodeBuffer.remaining()];
        decodeBuffer.get(decodedData);
        decodeBuffer.flip();
        return decodedData;
    }

    public void play() {
        int totalDecodedBytes = 0;
        try {
            SourceDataLine speaker = AudioSystem.getSourceDataLine(audioFormat);
            speaker.open();
            speaker.start();
            for (LogicalOggStream stream : (Collection<LogicalOggStream>) oggFile.getLogicalStreams()) {
                byte[] nextPacket = stream.getNextOggPacket();
                while (nextPacket != null) {
                    byte[] decodedData = decode(nextPacket);
                    if(decodedData != null) {
                        // Write packet to SourceDataLine
                        speaker.write(decodedData, 0, decodedData.length);
                        totalDecodedBytes += decodedData.length;
                    }
                    nextPacket = stream.getNextOggPacket();
                }
            }
            speaker.drain();
            speaker.close();
            System.out.println(String.format("Decoded to %d bytes", totalDecodedBytes));
        } catch (Exception e) {
            e.printStackTrace();
        }
    }
}

- Justin

---

有关这个问题有什么想法吗？“主线程中的异常java.lang.UnsatisfiedLinkError：在java.library.path中没有jnopus 在ClassLoader.java:1886中加载库 在Runtime.java:849中加载库0 在System.java:1088中加载库 在org.jitsi.impl.neomedia.codec.audio.opus.Opus.<clinit>(Opus.java:70)中 在OpusAudioPlayer.java:39中初始化(Opus.java:70) 在OpusAudioPlayer.java:30中的主函数(OpusAudioPlayer.java:30) - QGA

当您的构建路径中未包含jnopus库（平台相关）时，就会发生这种情况。检查您的构建并确保在libjitsi的“本地库依赖项”下包含所需的本地库。 - LiveMynd

根据我的经验：J-Ogg在使用FileStream从文件中读取数据时运行良好，但是当您尝试使用BasicStream从InputStream中读取数据时，它完全失效了。然而，VorbisJava没有同样的问题。 - MrPowerGamerBR

Answer 2

看了你的代码，我猜你误解了“帧长度”的含义。你正在取字节数，但帧长度直接取决于文件的编码方式。

以48000 Hz录制的音频文件每秒有48000个样本。这个音频样本通常是一个16位整数（2个字节），这意味着在非编码形式（PCM-WAV）下每秒会有48000 * 2个字节。

像opus这样的音频编码器会一次性采集多个音频样本并将它们编码成一个数据包。这就是帧。在48 kHz下，opus的值可以为120、240、480、960、1920和2880。

- David