NWaves is a .NET library for 1D signal processing focused specifically on audio processing.
NWaves is available on NuGet. Install it via
PM> Install-Package NWaves
New version 0.9.4 is out! Faster, smarter, more features. Read about changes here
- major DSP transforms (FFT, DCT, MDCT, STFT, FWT, Hilbert, Hartley, Mellin, cepstral, Goertzel)
- signal builders (sine, white/pink/red/Perlin noise, awgn, triangle, sawtooth, square, pulse, ramp, ADSR, wavetable)
- basic LTI digital filters (moving average, comb, Savitzky-Golay, pre/de-emphasis, DC removal, RASTA)
- FIR/IIR filtering (offline and online), zero-phase filtering
- BiQuad filters (low-pass, high-pass, band-pass, notch, all-pass, peaking, shelving)
- 1-pole filters (low-pass, high-pass)
- IIR filters (Bessel, Butterworth, Chebyshev I & II, Elliptic, Thiran)
- basic operations (convolution, cross-correlation, rectification, amplification)
- block convolution (overlap-add / overlap-save offline and online)
- basic filter design & analysis (group delay, zeros/poles, BP, BR, HP from/to LP, SOS, combining filters)
- state space representation of LTI filters
- FIR filter design: frequency sampling, window-sinc, equiripple (Remez / Parks-McClellan)
- non-linear filters (median filter, overdrive and distortion effects)
- windowing functions (Hamming, Blackman, Hann, Gaussian, Kaiser, KBD, triangular, Lanczos, flat-top, Bartlett-Hann)
- psychoacoustic filter banks (Mel, Bark, Critical Bands, ERB, octaves) and VTLN warping
- customizable feature extraction (time-domain, spectral, MFCC, PNCC/SPNCC, LPC, LPCC, PLP, AMS)
- preconfigured MFCC extractors: HTK (MFCC-FB24), Slaney (MFCC-FB40)
- LPC conversions: LPC<->cepstrum, LPC<->LSF
- feature post-processing (mean and variance normalization, adding deltas) and CSV serialization
- spectral features (centroid, spread, flatness, entropy, rolloff, contrast, crest, decrease, noiseness, MPEG7)
- harmonic features (harmonic centroid and spread, inharmonicity, tristimulus, odd-to-even ratio)
- time-domain characteristics (rms, energy, zero-crossing rate, entropy)
- pitch tracking (autocorrelation, YIN, ZCR + Schmitt trigger, HSS/HPS, cepstrum)
- time scale modification (phase vocoder, PV with identity phase locking, WSOLA, PaulStretch)
- simple resampling, interpolation, decimation
- bandlimited resampling
- wavelets: haar, db, symlet, coiflet
- polyphase filters
- noise reduction (spectral subtraction, sciPy-style Wiener filtering)
- envelope following
- sound effects (echo, tremolo, wahwah, phaser, chorus, vibrato, flanger, pitch shift, morphing, robotize, whisperize)
- harmonic/percussive separation
- Griffin-Lim algorithm
- Karplus-Strong synthesis
- PADSynth synthesis
- adaptive filtering (LMS, NLMS, LMF, SignLMS, RLS)
- simple modulation/demodulation (AM, ring, FM, PM)
- simple audio playback and recording
NWaves was initially intended for research, visualizing and teaching basics of DSP and sound programming. All algorithms are coded in C# as simple as possible and were first designed mostly for offline processing (now many online methods are also available). It doesn't mean, though, that the library could be used only in toy projects; yes, it's not written in C/C++ or Asm, but it's not that very slow for many purposes either.
// Create signal from samples repeated 100 times
float[] samples = new [] { 0.5f, 0.2f, -0.3f, 1.2f, 1.6f, -1.8f, 0.3f, -0.2f };
var s = new DiscreteSignal(8000, samples).Repeat(100);
var length = s.Length;
var duration = s.Duration;
var echoSignal = s + s.Delay(50);
var marginSignal = s.First(64).Concatenate(s.Last(64));
var repeatMiddle = s[400, 500].Repeat(10);
var mean = s.Samples.Average();
var sigma = s.Samples.Average(x => (x - mean) * (x - mean));
var normSignal = s - mean;
normSignal.Attenuate(sigma);
DiscreteSignal sinusoid =
new SineBuilder()
.SetParameter("frequency", 500.0/*Hz*/)
.SetParameter("phase", Math.PI / 6)
.OfLength(1000)
.SampledAt(44100/*Hz*/)
.Build();
DiscreteSignal noise =
new RedNoiseBuilder()
.SetParameter("min", -2.5)
.SetParameter("max", 2.5)
.OfLength(800)
.SampledAt(44100)
.DelayedBy(200)
.Build();
DiscreteSignal noisy =
new SineBuilder()
.SetParameter("min", -10.0)
.SetParameter("max", 10.0)
.SetParameter("freq", 1200.0/*Hz*/)
.OfLength(1000)
.SampledAt(44100)
.SuperimposedWith(noise)
.Build();
Signal builders can also act as real-time generators of samples:
SignalBuilder lfo =
new TriangleWaveBuilder()
.SetParameter("min", 100)
.SetParameter("max", 1500)
.SetParameter("frequency", 2.0/*Hz*/)
.SampledAt(16000/*Hz*/);
//while (...)
{
var sample = lfo.NextSample();
//...
}
DiscreteSignal left, right;
// load
using (var stream = new FileStream("sample.wav", FileMode.Open))
{
var waveFile = new WaveFile(stream);
left = waveFile[Channels.Left];
right = waveFile[Channels.Right];
}
// save
using (var stream = new FileStream("saved_mono.wav", FileMode.Create))
{
var waveFile = new WaveFile(left);
waveFile.SaveTo(stream);
}
using (var stream = new FileStream("saved_stereo.wav", FileMode.Create))
{
var waveFile = new WaveFile(new [] { left, right });
waveFile.SaveTo(stream);
}
For each transform there's a corresponding transformer object.
Each transformer object has Direct()
and Inverse()
methods.
// Complex FFT transformer:
var fft = new Fft(1024);
// Real FFT transformer (faster):
var rfft = new RealFft(1024);
float[] real = signal.First(1024).Samples;
float[] imag = new float [1024];
// in-place FFT
fft.Direct(real, imag);
// ...do something with real and imaginary parts of the spectrum...
// in-place IFFT
fft.Inverse(real, imag);
// post-processed FFT:
var magnitudeSpectrum =
fft.MagnitudeSpectrum(signal[1000, 2024]);
var powerSpectrum =
fft.PowerSpectrum(signal.First(1024), normalize: false);
var logPowerSpectrum =
fft.PowerSpectrum(signal.Last(1024))
.Samples
.Select(s => Scale.ToDecibel(s))
.ToArray();
// rfft will run faster;
// real FFT transforms real-valued signal to complex-valued spectrum,
// hence prototypes of methods are same, except Direct/Inverse:
rfft.Direct(real, real, imag); // real -> (real, imag)
rfft.Inverse(real, imag, real); // (real, imag) -> real
var magnitudeSpectrum =
rfft.MagnitudeSpectrum(signal[1000, 2024]);
var powerSpectrum =
rfft.PowerSpectrum(signal.First(1024), normalize: false);
// ...
Lot of methods in NWaves have overloaded versions with output buffers as parameters. So reuse memory whenever possible:
float[] spectrum = new float[1024];
for (var i = start; i < end; i += step)
{
rfft.MagnitudeSpectrum(signal[i, i + 1024], spectrum);
// ...
// do something with spectrum
}
// Short-Time Fourier Transform:
var stft = new Stft(1024, 256, WindowTypes.Hamming);
var timefreq = stft.Direct(signal);
var reconstructed = stft.Inverse(timefreq);
var spectrogram = stft.Spectrogram(signal);
// Cepstral transformer:
var ct = new CepstralTransform(24, fftSize: 512);
// complex cepstrum
var cepstrum = ct.Direct(signal);
// or
ct.Direct(input, output);
// real cepstrum
ct.RealCepstrum(input, output);
var fwt = new Fwt(192, new Wavelet("db5"));
// or
//var fwt = new Fwt(192, new Wavelet(WaveletFamily.Daubechies, 5));
var output = new float[192];
var reconstructed = new float[192];
fwt.Direct(input, output);
fwt.Inverse(output, reconstructed);
// convolution
var conv = Operation.Convolve(signal, kernel);
var xcorr = Operation.CrossCorrelate(signal1, signal2);
// block convolution
var filtered = Operation.BlockConvolve(signal, kernel, 4096, FilteringMethod.OverlapAdd);
// resampling
var resampled = Operation.Resample(signal, 22050);
var interpolated = Operation.Interpolate(signal, 3);
var decimated = Operation.Decimate(signal, 2);
var updown = Operation.ResampleUpDown(signal, 3, 2);
// time scale modification
var stretch = Operation.TimeStretch(signal, 0.7, TsmAlgorithm.Wsola);
var cool = Operation.TimeStretch(signal, 16, TsmAlgorithm.PaulStretch);
// envelope following
var envelope = Operation.Envelope(signal);
// rectification
var halfRect = Operation.HalfRectify(signal);
var fullRect = Operation.FullRectify(signal);
// spectral subtraction
var clean = Operation.SpectralSubtract(signal, noise);
var maFilter = new MovingAverageFilter(7);
var smoothedSignal = maFilter.ApplyTo(signal);
var frequency = 800.0/*Hz*/;
var notchFilter = new BiQuad.NotchFilter(frequency / signal.SamplingRate);
var notchedSignal = notchFilter.ApplyTo(signal);
// filter analysis:
var transferFunction = new TransferFunction(new [] { 1, 0.5, 0.2 }, new [] { 1, -0.8, 0.3 });
var filter = new IirFilter(transferFunction);
// we can also write this:
// var filter = new IirFilter(new [] { 1, 0.5, 0.2 }, new [] { 1, -0.8, 0.3 });
// var transferFunction = filter.Tf;
// ...
// if we simply want to apply filter and don't care much about FDA precision:
// read more in tutorial
var impulseResponse = transferFunction.ImpulseResponse();
var magnitudeResponse = transferFunction.FrequencyResponse().Magnitude;
var phaseResponse = transferFunction.FrequencyResponse().Phase;
var b = transferFunction.Numerator;
var a = transferFunction.Denominator;
var zeros = transferFunction.Zeros;
var poles = transferFunction.Poles;
var gd = transferFunction.GroupDelay();
var pd = transferFunction.PhaseDelay();
// some examples of FIR filter design:
var kernel = DesignFilter.FirWinLp(345, 0.15f);
var lpFilter = new FirFilter(kernel);
// HP filter can be obtained from LP with the same cutoff frequency:
var hpFilter = DesignFilter.FirLpToHp(lpFilter);
// design BP filter
var bpFilter = DesignFilter.FirWinBp(123, 0.05f, 0.15f);
// design equiripple HP filter
var bpFilter = DesignFilter.FirEquirippleHp(123, 0.34, 0.355, 0.05, 0.95);
// sequence of filters:
var cascade = filter * firFilter * notchFilter;
var filtered = cascade.ApplyTo(signal);
// equivalent to:
var filtered = filter.ApplyTo(signal);
filtered = firFilter.ApplyTo(filtered);
filtered = notchFilter.ApplyTo(filtered);
// parallel combination of filters:
var parallel = filter1 + filter2;
filtered = parallel.ApplyTo(signal);
// audio effects:
var flanger = new FlangerEffect(signal.SamplingRate);
var wahwah = new WahwahEffect(signal.SamplingRate, lfoFrequency: 2/*Hz*/);
var processed = wahwah.ApplyTo(flanger.ApplyTo(signal));
// this will create intermediate copy of the signal
// FilterChain is memory-efficient:
var filters = new FilterChain();
filters.Add(flanger);
filters.Add(wahwah);
processed = filters.ApplyTo(signal);
Online processing is supported by all classes that implement the IOnlineFilter
interface.
Currently, all LTI filters, FilterChain
class, block convolvers (OlaBlockConvolver
, OlsBlockConvolver
) and audio effects contain the Process(sample)
and Process(bufferIn, bufferOut)
methods responsible for online processing.
Simply process data sample after sample:
var outputSample = filter.Process(sample);
Or prepare necessary buffers (or just use them if they come from another part of your system):
float[] output;
...
void NewChunkAvailable(float[] chunk)
{
filter.Process(chunk, output);
}
// if input chunk shouldn't necessarily be preserved, it can be overwritten:
void NewChunkAvailable(float[] chunk)
{
filter.Process(chunk, chunk);
}
Block convolvers:
// Overlap-Add / Overlap-Save
FirFilter filter = new FirFilter(kernel);
var blockConvolver = OlaBlockConvolver.FromFilter(filter, 4096);
// processing loop:
// while new input sample is available
{
var outputSample = blockConvolver.Process(sample);
}
// or:
// while new input buffer is available
{
blockConvolver.Process(input, output);
}
See also OnlineDemoForm code.
Highly customizable feature extractors are available for offline and online processing (MFCC family, LPC, pitch and lot of others).
var mfccOptions = new MfccOptions
{
SamplingRate = signal.SamplingRate,
FeatureCount = 13,
FrameDuration = 0.032/*sec*/,
HopDuration = 0.015/*sec*/,
FilterBankSize = 26,
PreEmphasis = 0.97,
//...unspecified parameters will have default values
};
var mfccExtractor = new MfccExtractor(mfccOptions);
var mfccVectors = mfccExtractor.ComputeFrom(signal);
// serialize current config to JSON file:
using (var config = new FileStream("file.json", FileMode.Create))
{
config.SaveOptions(mfccOptions);
}
var lpcOptions = new LpcOptions
{
SamplingRate = signal.SamplingRate,
LpcOrder = 15
};
var lpcExtractor = new LpcExtractor(lpcOptions);
var lpcVectors = lpcExtractor.ParallelComputeFrom(signal);
var opts = new MultiFeatureOptions
{
SamplingRate = signal.SamplingRate,
FeatureList = "centroid, flatness, c1+c2+c3"
};
var spectralExtractor = new SpectralFeaturesExtractor(opts);
opts.FeatureList = "all";
var tdExtractor = new TimeDomainFeaturesExtractor(opts);
var vectors = FeaturePostProcessing.Join(
tdExtractor.ParallelComputeFrom(signal),
spectralExtractor.ParallelComputeFrom(signal));
// each vector will contain 1) all time-domain features (energy, rms, entropy, zcr)
// 2) specified spectral features
// open config from JSON file:
PnccOptions options;
using (var config = new FileStream("file.json", FileMode.Open))
{
options = config.LoadOptions<PnccOptions>();
}
var pnccExtractor = new PnccExtractor(pnccOptions);
var pnccVectors = pnccExtractor.ComputeFrom(signal, /*from*/1000, /*to*/60000 /*sample*/);
FeaturePostProcessing.NormalizeMean(pnccVectors);
// serialization
using (var csvFile = new FileStream("mfccs.csv", FileMode.Create))
{
var serializer = new CsvFeatureSerializer(mfccVectors);
await serializer.SerializeAsync(csvFile);
}
Pre-processing
In speech processing, pre-emphasis filters are often applied to signal before main processing.
There are 3 options to perform pre-emphasis filtering:
- Set pre-emphasis coefficient in constructor of a feature extractor
- Apply filter before processing and process filtered signal
- Filter signal in-place and process it
The first option is slightly slower, however it won't allocate extra memory and it won't mutate input signal (so, perhaps, it should be the choice by default). If preserving of the input signal is not required, then the third option is the best. If the input signal must be preserved and extra memory is not an issue, then the second approach is preferred (it'll be faster).
// option 1:
var opts = new MfccOptions
{
SamplingRate = signal.SamplingRate,
FeatureCount = 13,
PreEmphasis = 0.95
};
var mfccExtractor = new MfccExtractor(opts);
var mfccVectors = mfccExtractor.ComputeFrom(signal);
// option 2:
// ApplyTo() will create new signal (allocate new memory)
opts.PreEmphasis = 0;
mfccExtractor = new MfccExtractor(opts);
var pre = new PreEmphasisFilter(0.95);
var filtered = pre.ApplyTo(signal);
mfccVectors = mfccExtractor.ComputeFrom(filtered);
// option 3:
// process array or DiscreteSignal samples in-place:
for (var i = 0; i < signal.Length; i++)
{
signal[i] = pre.Process(signal[i]);
}
// or simply:
// pre.Process(signal.Samples, signal.Samples);
mfccVectors = mfccExtractor.ComputeFrom(signal);
MciAudioPlayer
and MciAudioRecorder
work only at Windows-side, since they use winmm.dll and MCI commands.
IAudioPlayer player = new MciAudioPlayer();
// play entire file
await player.PlayAsync("temp.wav");
// play file from 16000th sample to 32000th sample
await player.PlayAsync("temp.wav", 16000, 32000);
// ...in some event handler
player.Pause();
// ...in some event handler
player.Resume();
// ...in some event handler
player.Stop();
// recording
IAudioRecorder recorder = new MciAudioRecorder();
// ...in some event handler
recorder.StartRecording(16000);
// ...in some event handler
recorder.StopRecording("temp.wav");