indexing spectrogram object

Created by: shivarajkandhasamy

Below is a code showing issue with special case of index addressing of a spectrogram object. It seems the issue depends on how the spectrogram object is created.

import numpy as np
from gwpy import timeseries, spectrogram
from scipy import signal
sample_rate = 256
segment_duration = 10
no_of_segments = 4
N = segment_duration * sample_rate * no_of_segments
mu, sigma = 0, 1  # mean and standar deviation
data = np.random.default_rng().normal(mu, sigma, N)
data_start_time = 0
time_series_data = timeseries.TimeSeries(
            data, t0=data_start_time, sample_rate=sample_rate
        )
specgram_gwpy = time_series_data.spectrogram(segment_duration, fftlength=segment_duration, overlap=0)
print(specgram_gwpy)

Output looks

Spectrogram([[2.36786241e-04, 1.42787448e-03, 5.53594133e-03,
              ..., 1.78440578e-02, 3.41429717e-03,
              1.59868986e-03],
             [1.21973698e-02, 1.48942053e-02, 3.62360384e-03,
              ..., 3.30778092e-02, 1.94371388e-02,
              2.54025866e-03],
             [1.05670841e-03, 1.33569450e-02, 1.62157161e-03,
              ..., 2.51058318e-03, 8.15885644e-03,
              1.04887447e-03],
             [5.47444481e-07, 9.16911852e-04, 2.53900384e-03,
              ..., 3.27919955e-04, 5.59133760e-03,
              5.44161576e-03]]
            unit: 1 / Hz,
            name: None,
            epoch: 0.0,
            channel: None,
            x0: 0.0 s,
            dx: 10.0 s,
            xindex: [ 0. 10. 20. 30.] s,
            y0: 0.0 Hz,
            dy: 0.1 Hz,
            yindex: [0.000e+00 1.000e-01 2.000e-01 ... 1.278e+02 1.279e+02 1.280e+02] Hz)

Creating spectrogram object in another way,

overlap_factor = 0
f, t, Sxx = signal.spectrogram(
            time_series_data.data,
            fs=sample_rate,
            window="hann",
            nperseg= segment_duration* sample_rate,
            noverlap=overlap_factor * segment_duration * sample_rate,
            nfft=segment_duration * sample_rate,
            mode="complex",
            detrend=False,
            )
specgram_scipy = spectrogram.Spectrogram(
        np.real(2*np.conj(Sxx.T)*Sxx.T),
        times=t + time_series_data.t0.value - (segment_duration / 2),
        frequencies=f,  # - (fftlength / 2)
    )
print(specgram_scipy)

The output looks

Spectrogram([[0.00120062, 0.0018412 , 0.00553594, ...,
              0.01784406, 0.0034143 , 0.00319738],
             [0.02590223, 0.01547779, 0.0036236 , ...,
              0.03307781, 0.01943714, 0.00508052],
             [0.00469127, 0.01464505, 0.00162157, ...,
              0.00251058, 0.00815886, 0.00209775],
             [0.0013641 , 0.00231494, 0.002539  , ...,
              0.00032792, 0.00559134, 0.01088323]]
            unit: dimensionless,
            name: None,
            epoch: 0.0,
            channel: None,
            x0: 0.0 s,
            dx: 10.0 s,
            xindex: [ 0. 10. 20. 30.] s,
            y0: 0.0 Hz,
            dy: 0.1 Hz,
            yindex: [0.000e+00 1.000e-01 2.000e-01 ... 1.278e+02 1.279e+02 1.280e+02] Hz)

Now if I try to access a particular element using the below syntax I get error with specgram_scipy (the specgram_gwpy works fine).

ii=0
print(specgram_gwpy[ii:ii+1])
print(specgram_scipy[ii:ii+1])