Pywayne Plot

Enhanced spectrogram visualization tools for professional time-frequency analysis.

Quick Start

import matplotlib.pyplot as plt
from pywayne.plot import regist_projection, parula_map
import numpy as np

# Register custom projection
regist_projection()

# Create spectrogram
fig, ax = plt.subplots(subplot_kw={'projection': 'z_norm'})
spec, freqs, t, im = ax.specgram(
    x=signal_data,
    Fs=100,
    NFFT=128,
    noverlap=96,
    cmap=parula_map,
    scale='dB'
)
ax.set_ylabel('Frequency (Hz)')
plt.colorbar(im, label='Magnitude (dB)')
plt.show()

Functions

regist_projection

Register the custom SpecgramAxes projection. Must be called before using the enhanced specgram functionality.

from pywayne.plot import regist_projection
regist_projection()

SpecgramAxes.specgram

Enhanced spectrogram with advanced features.

Key Parameters:

Parameter	Description	Default
NFFT	FFT window length (points)	256
Fs	Sampling frequency (Hz)	2
noverlap	Overlap points between windows	128
cmap	Colormap (use parula_map)	-
mode	'psd', 'magnitude', 'angle', 'phase'	'psd'
scale	'dB' or 'linear'	'dB'
normalize	'global', 'local', 'none'	'global'
freq_scale	Frequency scaling factor	1.0
Fc	Center frequency offset (Hz)	0

Returns:

• spec - 2D spectrogram array (n_freqs, n_times)
• freqs - Frequency axis array
• t - Time axis array
• im - matplotlib image object (for colorbar)

get_specgram_params

Auto-recommend STFT parameters based on signal characteristics.

from pywayne.plot import get_specgram_params

params = get_specgram_params(
    signal_length=10000,
    sampling_rate=100,
    time_resolution=0.1  # or freq_resolution=0.5
)
# Returns: NFFT, noverlap, actual_freq_res, actual_time_res, n_segments

parula_map

MATLAB-style perceptually uniform colormap for scientific visualization.

from pywayne.plot import parula_map
plt.imshow(data, cmap=parula_map)

Usage Examples

IMU Signal Analysis

fs = 100  # Sampling rate
win_time, step_time = 1, 0.1

fig, ax = plt.subplots(subplot_kw={'projection': 'z_norm'})
spec, freqs, t, im = ax.specgram(
    x=acc_data,
    Fs=fs,
    NFFT=int(win_time * fs),
    noverlap=int((win_time - step_time) * fs),
    scale='dB',
    cmap=parula_map
)
ax.set_ylabel('Frequency (Hz)')
ax.set_ylim(0, 30)

Physiological Signals (PPG - Heart Rate)

# Convert Hz to bpm for heart rate visualization
fig, ax = plt.subplots(subplot_kw={'projection': 'z_norm'})
spec, freqs, t, im = ax.specgram(
    x=ppg_signal,
    Fs=100,
    NFFT=400,
    noverlap=300,
    freq_scale=60,  # Hz -> bpm
    scale='dB'
)
ax.set_ylabel('Heart Rate (bpm)')
ax.set_ylim(40, 180)

Vibration Analysis with Global Normalization

fig, ax = plt.subplots(subplot_kw={'projection': 'z_norm'})
spec, freqs, t, im = ax.specgram(
    x=vibration_data,
    Fs=1000,
    NFFT=1024,
    noverlap=512,
    scale='linear',
    normalize='global'
)
plt.colorbar(im, label='Normalized Magnitude')

High-Resolution Analysis with Zero-Padding

fig, ax = plt.subplots(subplot_kw={'projection': 'z_norm'})
spec, freqs, t, im = ax.specgram(
    x=signal,
    Fs=100,
    NFFT=100,
    pad_to=512,  # Zero-pad for smoother spectrum
    noverlap=80,
    scale='dB'
)

Scale and Normalization Modes

Scale Modes

Mode	Description	Use Case
dB	Logarithmic (10log10 for PSD, 20log10 for magnitude)	Large dynamic range signals
linear	Linear amplitude	Direct amplitude comparison

Normalization Modes (only for scale='linear')

Mode	Description	Use Case
global	Z/max(Z), preserves relative intensity	Compare intensity across time
local	Per-column normalization to [0,1]	Focus on frequency content over time
none	No normalization	Raw spectrogram values

Frequency Scaling

freq_scale	Unit	Use Case
1.0	Hz	Default, most signals
60	bpm	Heart rate, respiration rate
0.001	kHz	Audio signals

Example: freq_scale=60 converts 2 Hz → 120 bpm

Resolution Guidelines

• Frequency resolution: Δf = Fs / NFFT
• Time resolution: Δt = (NFFT - noverlap) / Fs
• Trade-off: Cannot simultaneously achieve high frequency and time resolution

Use get_specgram_params() to auto-calculate optimal parameters.

Interactive Analysis

spec, freqs, t, im = ax.specgram(...)

def on_click(event):
    if event.xdata and event.inaxes == ax:
        time_idx = np.argmin(np.abs(t - event.xdata))
        plt.figure()
        plt.plot(freqs, spec[:, time_idx])
        plt.title(f'FFT at t={event.xdata:.2f}s')
        plt.show()

fig.canvas.mpl_connect('button_press_event', on_click)

Application Areas

• IMU data: Accelerometer and gyroscope analysis
• Physiological signals: PPG (heart rate), ECG, respiration
• Vibration analysis: Machinery fault diagnosis
• Audio processing: Speech and audio spectrum analysis

Notes

• Always call regist_projection() before using projection='z_norm'
• parula_map is recommended for best perceptual uniformity
• dB mode automatically handles log(0) issues
• For better FFT efficiency, set NFFT to power of 2