Question

I’m working on an app that should do some audio signal processing. I need to measure the audio level in each one of the buffers I get (through the Callback function). I’ve been searching the web for some time, and I found that there is a build-in property called Current level metering:

AudioQueueGetProperty(recordState->queue,kAudioQueueProperty_CurrentLevelMeter,meters,&dlen);

This property gets me the average or peak audio level, but it’s not synchronised to the current buffer.
I figured out I need to calculate the audio level from the buffer data by myself, so I had this:

double calcAudioRMS (SInt16 * audioData, int numOfSamples)
{
double RMS, adPercent;
RMS = 0;

for (int i=0; i<numOfSamples; i++)
{
    adPercent=audioData[i]/32768.0f;
    RMS += adPercent*adPercent;
}
RMS = sqrt(RMS / numOfSamples);
return RMS;
}

This function gets the audio data (casted into Sint16) and the number of samples in the current buffer. The numbers I get are indeed between 0 and 1, but they seem to be rather random and low comparing to the numbers I got from the built-in audio level metering.

The recording audio format is:

format->mSampleRate = 8000.0;
format->mFormatID = kAudioFormatLinearPCM; 
format->mFramesPerPacket = 1;
format->mChannelsPerFrame = 1;
format->mBytesPerFrame = 2;
format->mBytesPerPacket = 2;
format->mBitsPerChannel = 16;
format->mReserved = 0;
format->mFormatFlags = kLinearPCMFormatFlagIsSignedInteger |kLinearPCMFormatFlagIsPacked;

My question is how to get the right values from the buffer? Is there a built-in function \ property for this? Or should I calculate the audio level myself, and how to do it?

Thanks in advance.

Answer 1

Your calculation for RMS power is correct. I’d be inclined to say that you have a fewer number of samples than Apple does, or something similar, and that would explain the difference. You can check by inputting a loud sine wave, and checking that Apple (and you) calculate RMS power at 1/sqrt(2).

Unless there’s a good reason, I would use Apple’s power calculations. I’ve used them, and they seem good to me. Additionally, generally you don’t want RMS power, you want RMS power as decibels, or use the kAudioQueueProperty_CurrentLevelMeterDB constant. (This depends on if you’re trying to build an audio meter, or truly display the audio power)