jono/selcal
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fft over a rolling sample window for greater bin precision

Browse Source
This commit is contained in:
Jono Targett 2024-05-26 15:34:57 +09:30
parent cf8a9dcfc1
commit 8aecf684f9
1 changed files with 27 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

47
scripts/selcal-fft.py
View File

@ -95,34 +95,41 @@ delta_f = sample_rate / segment_size
# Determine the bin range for desired frequency range (100 Hz to 2000 Hz) # Determine the bin range for desired frequency range (100 Hz to 2000 Hz)
high_bin = int(max_freq / delta_f) high_bin = int(max_freq / delta_f)
seg_off = int(sample_rate * 0.1)
act_segs = len(filtered_data) // seg_off
# Initialize a 2D array to store DFT results (magnitude spectrum) # Initialize a 2D array to store DFT results (magnitude spectrum)
# Only store the bins within the desired frequency range # Only store the bins within the desired frequency range
dft_results = np.zeros((num_segments, high_bin)) dft_results = np.zeros((act_segs, high_bin))
for i in range(num_segments): for i in range(act_segs):
start = i * segment_size end = (i+1) * seg_off
end = start + segment_size start = end - segment_size
segment = filtered_data[start:end]
# Step 4: Apply the DFT try:
dft_result = fft(segment) segment = filtered_data[start:end]
magnitudes = np.abs(dft_result) # Step 4: Apply the DFT
total_energy = np.sum(magnitudes ** 2) dft_result = fft(segment)
normalized_magnitudes = magnitudes / np.sqrt(total_energy)
normalized_magnitude = np.mean(normalized_magnitudes)
# Store the magnitude spectrum in the 2D array, only for the desired frequency range magnitudes = np.abs(dft_result)
dft_results[i, :] = normalized_magnitudes[:high_bin] total_energy = np.sum(magnitudes ** 2)
normalized_magnitudes = magnitudes / np.sqrt(total_energy)
normalized_magnitude = np.mean(normalized_magnitudes)
scores = [ # Store the magnitude spectrum in the 2D array, only for the desired frequency range
10 * np.log10(np.sum(normalized_magnitudes[int((f-w)/delta_f):int((f+w)/delta_f)])) dft_results[i, :] = normalized_magnitudes[:high_bin]
for f,w in zip(frequencies, widths)
]
codes = get_largest_two_indices(scores, 3.0) scores = [
if codes: 10 * np.log10(np.sum(normalized_magnitudes[int((f-w)/delta_f):int((f+w)/delta_f)]))
print([frequencies[code] for code in codes]) for f,w in zip(frequencies, widths)
]
codes = get_largest_two_indices(scores, 3.0)
if codes:
print([frequencies[code] for code in sorted(codes)])
except:
pass
# Step 5: Plot the spectrogram # Step 5: Plot the spectrogram