使用哪种方法实时读取音频样本

Question

For a particular project, I need to:

1. Access individual audio samples from the microphone or an audio file,
2. Extract a property from these values every 250 ms or so, and then
3. Display that property on-screen in ~real-time (up to 100 ms delay is fine).

I've looked at various approaches—Audio Units, Audio Graphs, AVAudioEngine, AudioTapProcessor—but not sure which is the right path for a Swift project aimed at iOS8 and iOS9 only. The AudioTapProcessor works well for accessing the audio samples from an audio file, but not sure about the mic input nor the Swift support.

Which approach best fits these requirements? Thanks for reading.

UPDATE: I went with AVAudioEngine and so far it's been a pretty great fit.

Answer 1

Audio Units and graphs would go hand and hand. Audio units are the components and the graph is the mechanism that connects them together. Using units and a graph will give you the best realtime (low latency) performance and options. I find that Objective C fits better with core audio since core audio was originally ac api.

I recently answered a question concerning ring buffers and used this project as a demo. This project plays a tone while it records from the mic and allows you to process by reading the latest samples from a ring. This may be a good starting point. You can remove the tone playing if needed.

Answer 2

I believe that a minimalistic low-level approach with a single audio component of kAudioUnitSubType_RemoteIO would perform reliably in terms of low latency and Swift support. Presuming that the interface (named here for convenience) myAudioController is properly declared and initialized, the following code inside the registered render callback should do the real-time I/O mapping (here written in C , though):

myAudioController *myController = (myAudioController *)inRefCon;

//this would render inNumberFrames of data from audio input...
AudioUnitRender(myController->audioUnit,
                            ioActionFlags,
                            inTimeStamp,
                            bus1,
                            inNumberFrames,
                            ioData);

//from here on, individual samples can be monitored and processed...                           
AudioBuffer buffer = ioData->mBuffers[0]; 

The equivalent code snippet in Swift would probably look like this:

let myController = UnsafeMutablePointer<myAudioController>(inRefCon).memory
AudioUnitRender(myController.audioUnit, 
                            ioActionFlags, 
                            inTimeStamp, 
                            bus1, 
                            inNumberFrames, 
                            ioData)

for buffer in UnsafeMutableAudioBufferListPointer(ioData) { ... }

Please feel free to consult Apple's reference page for details - it is very well documented: https://developer.apple.com/library/ios/documentation/AudioUnit/Reference/AUComponentServicesReference/#//apple_ref/c/func/AudioUnitRender

This might also be a valuable site, with C examples from a must-read textbook re-written in Swift : https://github.com/AlesTsurko/LearningCoreAudioWithSwift2.0

Important is understanding what you are doing. Everything else should be pretty much self-explanatory and should not involve too much work. Hope this can help…