How to encode and decode Real-time Audio using OpusCodec in IOS?

烂漫一生 提交于 2021-02-07 07:53:41

问题


I am working on a app which has following requirements :

  1. Record real time audio from iOS device (iPhone)
  2. Encode this audio data to Opus data and send it to server over WebSocket
  3. Decode received data to pcm again
  4. Play received audio from WebSocket server on iOS device(iPhone)

I've used AVAudioEngine for this.

 var engine = AVAudioEngine()
 var input: AVAudioInputNode = engine.inputNode
 var format: AVAudioFormat = input.outputFormat(forBus: AVAudioNodeBus(0))
 input.installTap(onBus: AVAudioNodeBus(0), bufferSize: AVAudioFrameCount(8192), format: format, block: { buf, when in
 // ‘buf' contains audio captured from input node at time 'when'
 })

 // start engine

And I converted AVAudioPCMBuffer to Data using this function

func toData(PCMBuffer: AVAudioPCMBuffer) -> Data {
    let channelCount = 1
    let channels = UnsafeBufferPointer(start: PCMBuffer.floatChannelData, count: channelCount)
    let ch0Data = NSData(bytes: channels[0], length:Int(PCMBuffer.frameLength * PCMBuffer.format.streamDescription.pointee.mBytesPerFrame))
    return ch0Data as Data
}

I've found Opus Library from CocoaPod libopus libopus

I have searched a lot on how to use OpusCodec in IOS but haven't got the solution.

How to encode and decode this data using OpusCodec ? And Am I need jitterBuffer ? If I need How to use it in IOS

This Code for Opus Codec but voice doesn't clear

#import "OpusManager.h"
#import <opus/opus.h>

#define SAMPLE_RATE 16000
#define CHANNELS 1
#define BITRATE SAMPLE_RATE * CHANNELS
/**
* Audio frame size
* It is divided by time. When calling, you must use the audio data of 
exactly one frame (multiple of 2.5ms: 2.5, 5, 10, 20, 40, 60ms).
* Fs/ms   2.5     5       10      20      40      60
* 8kHz    20      40      80      160     320     480
* 16kHz   40      80      160     320     640     960
* 24KHz   60      120     240     480     960     1440
* 48kHz   120     240     480     960     1920    2880
*/
#define FRAME_SIZE 320

#define APPLICATION         OPUS_APPLICATION_VOIP
#define MAX_PACKET_BYTES    (FRAME_SIZE * CHANNELS * sizeof(float))
#define MAX_FRAME_SIZE      (FRAME_SIZE * CHANNELS * sizeof(float))

typedef opus_int16 OPUS_DATA_SIZE_T;

@implementation OpusManager {
    OpusEncoder *_encoder;
    OpusDecoder *_decoder;
}

int size;
int error;
unsigned char encodedPacket[MAX_PACKET_BYTES];

- (instancetype)init {
    self = [super init];
    if (self) {

        size = opus_encoder_get_size(CHANNELS);
        _encoder = malloc(size);
        error = opus_encoder_init(_encoder, SAMPLE_RATE, CHANNELS, APPLICATION);   
        _encoder = opus_encoder_create(SAMPLE_RATE, CHANNELS, APPLICATION, &error);
        _decoder = opus_decoder_create(SAMPLE_RATE, CHANNELS, &error);

        opus_encoder_ctl(_encoder, OPUS_SET_BITRATE(BITRATE));
        opus_encoder_ctl(_encoder, OPUS_SET_COMPLEXITY(10));
        opus_encoder_ctl(_encoder, OPUS_SET_SIGNAL(OPUS_SIGNAL_VOICE));
        opus_encoder_ctl(_encoder, OPUS_SET_VBR(0));
        opus_encoder_ctl(_encoder, OPUS_SET_APPLICATION(APPLICATION));
        opus_encoder_ctl(_encoder, OPUS_SET_DTX(1));
        opus_encoder_ctl(_encoder, OPUS_SET_INBAND_FEC(0));
        opus_encoder_ctl(_encoder, OPUS_SET_BANDWIDTH(12000));
        opus_encoder_ctl(_encoder, OPUS_SET_PACKET_LOSS_PERC(1));
        opus_encoder_ctl(_encoder, OPUS_SET_INBAND_FEC(1));
        opus_encoder_ctl(_encoder, OPUS_SET_FORCE_CHANNELS(CHANNELS));
        opus_encoder_ctl(_encoder, OPUS_SET_PACKET_LOSS_PERC(1));
     }
     return self;
}

- (NSData *)encode:(NSData *)PCM {

    opus_int16 *PCMPtr = (opus_int16 *)PCM.bytes;
    int PCMSize = (int)PCM.length / sizeof(opus_int16);
    opus_int16 *PCMEnd = PCMPtr + PCMSize;
    NSMutableData *mutData = [NSMutableData data];
    unsigned char encodedPacket[MAX_PACKET_BYTES];

    // Record opus block size
    OPUS_DATA_SIZE_T encodedBytes = 0;

    while (PCMPtr + FRAME_SIZE < PCMEnd) {
    encodedBytes = opus_encode_float(_encoder, (const float *) PCMPtr, FRAME_SIZE, encodedPacket, MAX_PACKET_BYTES);

    if (encodedBytes <= 0) {
        NSLog(@"ERROR: encodedBytes<=0");
        return nil;
    }
    NSLog(@"encodedBytes: %d",  encodedBytes);

    // Save the opus block size
    [mutData appendBytes:&encodedBytes length:sizeof(encodedBytes)];

    // Save opus data
    [mutData appendBytes:encodedPacket length:encodedBytes];

    PCMPtr += FRAME_SIZE;
    }

    NSLog(@"mutData: %lu", (unsigned long)mutData.length);
    NSLog(@"encodedPacket: %s", encodedPacket);

    return mutData.length > 0 ? mutData : nil;

}

- (NSData *)decode:(NSData *)opus {

    unsigned char *opusPtr = (unsigned char *)opus.bytes;
    int opusSize = (int)opus.length;
    unsigned char *opusEnd = opusPtr + opusSize;

    NSMutableData *mutData = [NSMutableData data];

    float decodedPacket[MAX_FRAME_SIZE];
    int decodedSamples = 0;

    // Save data for opus block size
    OPUS_DATA_SIZE_T nBytes = 0;

    while (opusPtr < opusEnd) {
        // Take out the opus block size data
        nBytes = *(OPUS_DATA_SIZE_T *)opusPtr;
        opusPtr += sizeof(nBytes);

        decodedSamples = opus_decode_float(_decoder, opusPtr, nBytes,decodedPacket, MAX_FRAME_SIZE, 0);

        if (decodedSamples <= 0) {
            NSLog(@"ERROR: decodedSamples<=0");
            return nil;
        }
        NSLog(@"decodedSamples:%d", decodedSamples);
        [mutData appendBytes:decodedPacket length:decodedSamples *sizeof(opus_int16)];

        opusPtr += nBytes;
    }
    NSLog(@"mutData: %lu", (unsigned long)mutData.length);
    return mutData.length > 0 ? mutData : nil;
}

@end

回答1:


Try to lower the bandwidth or set an higher bitrate. I think 16kbit for a 12kHz bandwidth mono audio is probably too low. Think it will be better to leave bandwidth to auto with application VOIP setted. There could be other problems around but the "doesn't sound good" is not enough to analyze.



来源:https://stackoverflow.com/questions/55692517/how-to-encode-and-decode-real-time-audio-using-opuscodec-in-ios

易学教程内所有资源均来自网络或用户发布的内容,如有违反法律规定的内容欢迎反馈
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!