Added the sound recognition part

[vladturcuman]

Changed the MicroBitAudioProcessor to also be a DataSource. Added a Hann Window and Harmonic Product Spectrum before and after fft to get more accurate results for square wave detection.

Added MicroBitSoundRecogniser as a DataSink, which recognises given patterns in the frequency data. It connects to the audio processor to get the frequency analysis. The constructor is protected such that the class becomes abstract. EmojiRecogniser inherits it and implements the constructor - adds the happy sound as the sound to be recognised.

To recognise the happy sound the following main.cpp can be used:

#include "MicroBitSerial.h"
#include "MicroBit.h"
#include "CodalDmesg.h"
#include "MicroBitAudioProcessor.h"
#include "EmojiRecogniser.h"
#include "StreamNormalizer.h"
#include "Tests.h"

#include "LevelDetector.h"
#include "StreamSplitter.h"

static NRF52ADCChannel *mic = NULL;
static StreamNormalizer *processor = NULL;
static MicroBitAudioProcessor *fft = NULL;

static LevelDetector *level = NULL;
static StreamSplitter *splitter = NULL;

MicroBitSoundRecogniser *recogniser = NULL;

MicroBit uBit;

void onSound(ManagedString sound) {
    recogniser->stopAnalisying();

    uBit.display.scroll(sound);

    recogniser->startAnalisying(onSound);
}

int main()
{
    uBit.init();

    NRF52ADCChannel *mic = uBit.adc.getChannel(uBit.io.microphone);
    mic->setGain(7,0);
    uBit.io.runmic.setDigitalValue(1);
    uBit.io.runmic.setHighDrive(true);

    StreamNormalizer *processor = new StreamNormalizer(mic->output, 1.2f, true, DATASTREAM_FORMAT_8BIT_SIGNED, 10);
    StreamSplitter *splitter = new StreamSplitter(processor->output);
    fft         = new MicroBitAudioProcessor(*splitter);
    recogniser  = new EmojiRecogniser(*fft, uBit);

    recogniser->startAnalisying(onSound);

    while(1){
        uBit.display.print("-");
        uBit.sleep(1000);
    }

}

[vladturcuman]

Moved the pull request to the lancaster-university / codal-microbit-v2 repo, here.