Introduction

pymultitaper is a fast and easy-to-use small package for multitaper spectrogram/spectrum calculation, as well as oridnary (single-taper) spectrogram calculation.

Installation

Install via pip:

pip install pymultitaper

GPU computation is supported in a pre-release version. To install the GPU-enabled version, use:

pip install --pre pymultitaper

GPU support requires CuPy. For installation instructions, see the CuPy documentation.

Note: Cupy is an optional dependency. If you do not have CuPy installed, pymultitaper will still work using the CPU backend.

Usage

>>> from pymultitaper import multitaper_spectrogram, plot_spectrogram
>>> from scipy.io import wavfile
>>> fs, data = wavfile.read('test.wav')
>>> freqs,times,psd = multitaper_spectrogram(
...     data, fs,time_step=0.001,window_length=0.005,NW=4
... )
>>> fig,ax = plot_spectrogram(times,freqs,psd,cmap="viridis")

GPU Support

GPU usage is automatic: if you pass a CuPy array into the spectrogram functions, pymultitaper will use the GPU backend. NumPy arrays still use the CPU backend.

>>> import cupy as cp
>>> from pymultitaper import multitaper_spectrogram
>>> from scipy.io import wavfile
>>> fs, data = wavfile.read("test.wav")
>>> data_gpu = cp.asarray(data)
>>> freqs, times, psd = multitaper_spectrogram(
...     data_gpu, fs, time_step=0.001, window_length=0.005, NW=4
... )
>>> psd.shape
(len(freqs), len(times))

The returned arrays stay on the GPU until you explicitly move them back to NumPy, for example with cp.asnumpy(...), before plotting.

Examples