CODEWITHSUNDEEP
pythonscipyfftspectrogram
Shamoon
Shamoon

Reputation: 43619

When doing a scipy.signal.spectrogram, how can I tell how many time segments there will be?

I have:

    frequencies, times, spectrogram = signal.spectrogram(
        samples, sample_rate, nperseg=nperseg, window=signal.hann(nperseg), noverlap=noverlap, mode='magnitude')

I have nperseg=320 and noverlap=80, with a sample_rate=16000.

I had 232800 samples.

However, my spectrogram.shape is now (161, 969). So that's 161 frequency bins with 969 time segments. How was that 969 calculated?

Upvotes: 0

Views: 1447

Answers (1)

Shamoon
Shamoon

Reputation: 43619

    time_per_segment_ms = 20
    nperseg = int(sample_rate * 0.001 * time_per_segment_ms)
    overlap = nperseg // 4

    seconds_per_segment = (nperseg - overlap) / sample_rate
    ms_per_segment = int(seconds_per_segment * 1000)

That about does it

Upvotes: 1

Related Questions