DocsSignal Processing

Signal Processing Tools

GPU-accelerated signal processing tools for audio, time series, and sensor data analysis.

Available Tools

Tool	Description
FFT Analysis	Fast Fourier Transform
Spectrogram	Time-frequency analysis
Filter Designer	Design digital filters
Convolution	Signal convolution
Wavelet Transform	Wavelet analysis

FFT Analysis

Transform time-domain signals to frequency domain for spectral analysis.

import cyxwiz.signal as sig

# Forward FFT
frequencies = sig.fft(signal)

# Inverse FFT
reconstructed = sig.ifft(frequencies)

# Power spectrum
power = sig.power_spectrum(signal, sample_rate=44100)

# Magnitude and phase
magnitude = sig.magnitude(frequencies)
phase = sig.phase(frequencies)

Spectrogram

Time-frequency representation using Short-Time Fourier Transform (STFT).

# Compute spectrogram
spec = sig.spectrogram(
    signal,
    sample_rate=44100,
    window_size=1024,
    hop_size=256,
    window='hann'
)

# Mel spectrogram (for audio)
mel_spec = sig.mel_spectrogram(
    signal,
    sample_rate=44100,
    n_mels=128
)

# Log-mel spectrogram
log_mel = sig.log_mel_spectrogram(signal, sample_rate=44100)

Digital Filters

Filter Types

Lowpass: Remove high frequencies
Highpass: Remove low frequencies
Bandpass: Keep frequency range
Bandstop: Remove frequency range

Usage

# Design Butterworth lowpass
b, a = sig.butter(
    order=4,
    cutoff=1000,
    fs=44100,
    btype='low'
)

# Apply filter
filtered = sig.filtfilt(b, a, signal)

Wavelet Transform

# Continuous Wavelet Transform
cwt_result = sig.cwt(signal, wavelet='morlet', scales=range(1, 128))

# Discrete Wavelet Transform
cA, cD = sig.dwt(signal, wavelet='db4')

# Multi-level decomposition
coeffs = sig.wavedec(signal, wavelet='db4', level=5)

# Reconstruction
reconstructed = sig.waverec(coeffs, wavelet='db4')

Node Editor Integration

Node	Inputs	Outputs
FFT	Signal tensor	Frequency tensor
Spectrogram	Signal, params	2D spectrogram
Filter	Signal, coefficients	Filtered signal
Wavelet	Signal, wavelet type	Coefficients