Audio Processing: Playing with volume level

问题

I want to read a sound file from application bundle, copy it, play with its maximum volume level(Gain value or peak power, I'm not sure about the technical name of it), and then write it as another file to the bundle again.

I did the copying and writing part. Resulting file is identical to input file. I use AudioFileReadBytes() and AudioFileWriteBytes() functions of AudioFile services in AudioToolbox framework to do that.

So, I have the input file's bytes and also its audio data format(via use of AudioFileGetProperty() with kAudioFilePropertyDataFormat) but I can't find a variable in these to play with the original file's maximum volume level.

To clarify my purpose, I'm trying to produce another sound file of which volume level is increased or decreased relative to the original one, so I don't care about the system's volume level which is set by the user or iOS.

Is that possible to do with the framework I mentioned? If not, are there any alternative suggestions?

Thanks

---

edit: Walking through Sam's answer regarding some audio basics, I decided to expand the question with another alternative.

Can I use AudioQueue services to record existing sound file(which is in the bundle) to another file and play with the volume level(with the help of framework) during the recording phase?

---

update: Here's how I'm reading the input file and writing the output. Below code lowers the sound level for "some" of the amplitude values but with lots of noise. Interestingly, if I choose 0.5 as amplitude value it increases the sound level instead of lowering it, but when I use 0.1 as amplitude value it lowers the sound. Both cases involve disturbing noise. I think that's why Art is talking about normalization, but I've no idea about normalization.

AudioFileID inFileID;

CFURLRef inURL = [self inSoundURL];

AudioFileOpenURL(inURL, kAudioFileReadPermission, kAudioFileWAVEType, &inFileID)

UInt32 fileSize = [self audioFileSize:inFileID];
Float32 *inData = malloc(fileSize * sizeof(Float32)); //I used Float32 type with jv42's suggestion
AudioFileReadBytes(inFileID, false, 0, &fileSize, inData);

Float32 *outData = malloc(fileSize * sizeof(Float32));

//Art's suggestion, if I've correctly understood him

float ampScale = 0.5f; //this will reduce the 'volume' by -6db
for (int i = 0; i < fileSize; i++) {
    outData[i] = (Float32)(inData[i] * ampScale);
}

AudioStreamBasicDescription outDataFormat = {0};
[self audioDataFormat:inFileID];

AudioFileID outFileID;

CFURLRef outURL = [self outSoundURL];
AudioFileCreateWithURL(outURL, kAudioFileWAVEType, &outDataFormat, kAudioFileFlags_EraseFile, &outFileID)

AudioFileWriteBytes(outFileID, false, 0, &fileSize, outData);

AudioFileClose(outFileID);
AudioFileClose(inFileID);

回答1:

You won't find amplitude scaling operations in (Ext)AudioFile, because it's about the simplest DSP you can do.

Let's assume you use ExtAudioFile to convert whatever you read into 32-bit floats. To change the amplitude, you simply multiply:

float ampScale = 0.5f; //this will reduce the 'volume' by -6db
for (int ii=0; ii<numSamples; ++ii) {
    *sampOut = *sampIn * ampScale;
    sampOut++; sampIn++;
}

To increase the gain, you simply use a scale > 1.f. For example, an ampScale of 2.f would give you +6dB of gain.

If you want to normalize, you have to make two passes over the audio: One to determine the sample with the greatest amplitude. Then another to actually apply your computed gain.

Using AudioQueue services just to get access to the volume property is serious, serious overkill.

UPDATE:

In your updated code, you're multiplying each byte by 0.5 instead of each sample. Here's a quick-and-dirty fix for your code, but see my notes below. I wouldn't do what you're doing.

...

// create short pointers to our byte data
int16_t *inDataShort = (int16_t *)inData;
int16_t *outDataShort = (int16_t *)inData;

int16_t ampScale = 2;
for (int i = 0; i < fileSize; i++) {
    outDataShort[i] = inDataShort[i] / ampScale;
}

...

Of course, this isn't the best way to do things: It assumes your file is little-endian 16-bit signed linear PCM. (Most WAV files are, but not AIFF, m4a, mp3, etc.) I'd use the ExtAudioFile API instead of the AudioFile API as this will convert any format you're reading into whatever format you want to work with in code. Usually the simplest thing to do is read your samples in as 32-bit float. Here's an example of your code using ExtAudioAPI to handle any input file format, including stereo v. mono

void ScaleAudioFileAmplitude(NSURL *theURL, float ampScale) {
    OSStatus err = noErr;

    ExtAudioFileRef audiofile;
    ExtAudioFileOpenURL((CFURLRef)theURL, &audiofile);
    assert(audiofile);

    // get some info about the file's format.
    AudioStreamBasicDescription fileFormat;
    UInt32 size = sizeof(fileFormat);
    err = ExtAudioFileGetProperty(audiofile, kExtAudioFileProperty_FileDataFormat, &size, &fileFormat);

    // we'll need to know what type of file it is later when we write 
    AudioFileID aFile;
    size = sizeof(aFile);
    err = ExtAudioFileGetProperty(audiofile, kExtAudioFileProperty_AudioFile, &size, &aFile);
    AudioFileTypeID fileType;
    size = sizeof(fileType);
    err = AudioFileGetProperty(aFile, kAudioFilePropertyFileFormat, &size, &fileType);


    // tell the ExtAudioFile API what format we want samples back in
    AudioStreamBasicDescription clientFormat;
    bzero(&clientFormat, sizeof(clientFormat));
    clientFormat.mChannelsPerFrame = fileFormat.mChannelsPerFrame;
    clientFormat.mBytesPerFrame = 4;
    clientFormat.mBytesPerPacket = clientFormat.mBytesPerFrame;
    clientFormat.mFramesPerPacket = 1;
    clientFormat.mBitsPerChannel = 32;
    clientFormat.mFormatID = kAudioFormatLinearPCM;
    clientFormat.mSampleRate = fileFormat.mSampleRate;
    clientFormat.mFormatFlags = kLinearPCMFormatFlagIsFloat | kAudioFormatFlagIsNonInterleaved;
    err = ExtAudioFileSetProperty(audiofile, kExtAudioFileProperty_ClientDataFormat, sizeof(clientFormat), &clientFormat);

    // find out how many frames we need to read
    SInt64 numFrames = 0;
    size = sizeof(numFrames);
    err = ExtAudioFileGetProperty(audiofile, kExtAudioFileProperty_FileLengthFrames, &size, &numFrames);

    // create the buffers for reading in data
    AudioBufferList *bufferList = malloc(sizeof(AudioBufferList) + sizeof(AudioBuffer) * (clientFormat.mChannelsPerFrame - 1));
    bufferList->mNumberBuffers = clientFormat.mChannelsPerFrame;
    for (int ii=0; ii < bufferList->mNumberBuffers; ++ii) {
        bufferList->mBuffers[ii].mDataByteSize = sizeof(float) * numFrames;
        bufferList->mBuffers[ii].mNumberChannels = 1;
        bufferList->mBuffers[ii].mData = malloc(bufferList->mBuffers[ii].mDataByteSize);
    }

    // read in the data
    UInt32 rFrames = (UInt32)numFrames;
    err = ExtAudioFileRead(audiofile, &rFrames, bufferList);

    // close the file
    err = ExtAudioFileDispose(audiofile);

    // process the audio
    for (int ii=0; ii < bufferList->mNumberBuffers; ++ii) {
        float *fBuf = (float *)bufferList->mBuffers[ii].mData;
        for (int jj=0; jj < rFrames; ++jj) {
            *fBuf = *fBuf * ampScale;
            fBuf++;
        }
    }

    // open the file for writing
    err = ExtAudioFileCreateWithURL((CFURLRef)theURL, fileType, &fileFormat, NULL, kAudioFileFlags_EraseFile, &audiofile);

    // tell the ExtAudioFile API what format we'll be sending samples in
    err = ExtAudioFileSetProperty(audiofile, kExtAudioFileProperty_ClientDataFormat, sizeof(clientFormat), &clientFormat);

    // write the data
    err = ExtAudioFileWrite(audiofile, rFrames, bufferList);

    // close the file
    ExtAudioFileDispose(audiofile);

    // destroy the buffers
    for (int ii=0; ii < bufferList->mNumberBuffers; ++ii) {
        free(bufferList->mBuffers[ii].mData);
    }
    free(bufferList);
    bufferList = NULL;

}

回答2:

I think you should avoid working with 8 bits unsigned chars for audio, if you can. Try to get the data as 16 bits or 32 bits, that would avoid some noise/bad quality issues.

回答3:

For most common audio file formats there isn't a single master volume variable. Instead you will need to take (or convert to) the PCM sound samples and perform at least some minimal digital signal processing (multiply, saturate/limit/AGC, quantization noise shaping, and etc.) on each sample.

回答4:

If the sound file is normalized, there's nothing you can do to make the file louder. Except in the case of poorly encoded audio, volume is almost entirely the realm of the playback engine.

http://en.wikipedia.org/wiki/Audio_bit_depth

Properly stored audio files will have peak volume at or near the maximum value available for the file's bit depth. If you attempt to 'decrease the volume' of a sound file, you'll essentially just be degrading the sound quality.

来源：https://stackoverflow.com/questions/3982723/audio-processing-playing-with-volume-level