public override VowelType Decide(ReadOnlySpan <float> samples, int sampleRate)
{
var input = new MfccAccord(sampleRate, samples.Length, 0, 8000, 24).ComputeMfcc12D(samples);
this._network.Compute(input).Max(out var i);
return((VowelType)i);
}
private static void MelFilteredSpectrum()
{
const int windowSize = 2048;
const int sampleRate = 44100;
var mfccComputer = new MfccAccord(sampleRate, windowSize, 0, 8000, 8);
var hzAxis = mfccComputer.HzAxisOfMelSpectrum();
float[] ReadSamples(string fileName, double secs)
{
float[] xs;
using (var wavReader = new WaveFileReader(Path.Combine(CommonUtils.GetTrainingDataDirectory(), fileName)))
{
var provider = wavReader.ToSampleProvider().Skip(TimeSpan.FromSeconds(secs)).ToMono();
if (provider.WaveFormat.SampleRate != sampleRate)
{
throw new Exception();
}
xs = new float[windowSize];
for (var readSamples = 0; readSamples < xs.Length;)
{
var count = provider.Read(xs, readSamples, xs.Length - readSamples);
if (count == 0)
{
break;
}
readSamples += count;
}
}
return(xs);
}
void ShowSpectrum(string title, params (string, double)[] inputs)
private static void Mfcc()
{
const int windowSize = 2048;
int rate;
var data = new float[windowSize];
// 2.5s のところから 2048 サンプル取得してくる
using (var reader = new WaveFileReader(Path.Combine(CommonUtils.GetTrainingDataDirectory(), "あいうえお 2017-12-18 00-17-09.wav")))
{
var provider = reader.ToSampleProvider()
.Skip(TimeSpan.FromSeconds(2.5))
.ToMono();
rate = provider.WaveFormat.SampleRate;
for (var readSamples = 0; readSamples < data.Length;)
{
var delta = provider.Read(data, readSamples, data.Length - readSamples);
if (delta == 0)
{
throw new EndOfStreamException();
}
readSamples += delta;
}
}
var mfcc = new MfccAccord(rate, windowSize, 0, 8000, 24);
foreach (var x in mfcc.ComputeMfcc12D(data))
{
Console.WriteLine(x);
}
}
public override Task AddTrainingDataAsync(string csvFileName)
{
float[] samples;
int rate;
// 同じファイル名の .wav ファイルをロード
using (var wavReader = new WaveFileReader(Path.ChangeExtension(csvFileName, ".wav")))
{
var provider = wavReader.ToSampleProvider().ToMono();
rate = provider.WaveFormat.SampleRate;
// 全サンプル読み込んじゃえ
samples = new float[wavReader.SampleCount];
for (var readSamples = 0; readSamples < samples.Length;)
{
var count = provider.Read(samples, readSamples, samples.Length - readSamples);
if (count == 0)
{
break;
}
readSamples += count;
}
}
var tasks = new List <Task>();
using (var csvReader = new CsvReader(csvFileName, true))
{
const int windowSize = 2048;
var mfcc = new MfccAccord(rate, windowSize, 0, 6000, 8);
while (csvReader.ReadNextRecord())
{
var time = double.Parse(csvReader["Time"], CultureInfo.InvariantCulture);
var vowelType = ParseVowelType(csvReader["Class"]);
// どうしよう
if (vowelType == VowelType.Other)
{
continue;
}
// 並列でばんばか投げていくぞ
tasks.Add(Task.Run(() =>
{
var v = mfcc.MelSpectrum(new ReadOnlySpan <float>(samples, (int)(time * rate), windowSize));
v.Subtract(v.Mean(), v);
lock (this.TrainingData)
this.TrainingData.Add((v, vowelType));
}));
}
}
return(Task.WhenAll(tasks));
}
public override VowelType Decide(ReadOnlySpan <float> samples, int sampleRate)
{
var input = new MfccAccord(sampleRate, samples.Length, 0, 6000, 8).MelSpectrum(samples);
return((VowelType)this._teacher.Model.Decide(input));
}
public override VowelType Decide(ReadOnlySpan <float> samples, int sampleRate)
{
var input = new MfccAccord(sampleRate, samples.Length, 0, 8000, 24).ComputeMfcc12D(samples);
return((VowelType)this._teacher.Model.Decide(input));
}
