Very rough initial analysis of some sound recordings to identify whale songs

whalesong.py

from pathlib import Path
from scipy.io import wavfile
from scipy.fft import fft, fftfreq
import matplotlib.pyplot as plt
import numpy as np

path = Path('.') # path to whale song files

files = list(path.glob('*.wav'))

# use every nth sample, to speed up processing
# NOTE: divides maximum detectable frequency
downsample = 20

fig, ax = plt.subplots()

for file in files:
    sample_rate, data = wavfile.read(file)
    # basically copy scipy 1D Fourier transform example
    N = len(data) // downsample
    T = downsample / sample_rate
    x = np.linspace(0, N / sample_rate, N, endpoint=False)
    y = data[::downsample]
    valid_frequencies = N // 2 # Nyquist limit
    yf = 2 / N * np.abs(fft(y))[:valid_frequencies]
    xf = fftfreq(N, T)[:valid_frequencies]
    ax.loglog(xf, yf) # plot power spectrum in log-log form

ax.set_xlabel('frequency [Hz]')
plt.show()

Initial plots here: https://imgur.com/a/bLzsO3n