Wrong axis for bin deletion in _spectrum_smoothing

[Antoine101]

Hi,

I tried to calculate the Tone-to-Noise ratio on a 200ms snapshot of a stationnary signal.
I get the following error:

 File ~\Anaconda3\lib\site-packages\spyder_kernels\py3compat.py:356 in compat_exec
    exec(code, globals, locals)

  File d:\dev\alarm_new\identification\machine learning\postprocess_dataset.py:27
    dataset_sliced = load_slice_compute(

  File d:\dev\ml_tools\ml_tools\audio\machine_learning\load_slice_compute.py:116 in load_slice_compute
    feature_vector.compute(snapshot_norm, sampling_rate, filter_signal, low, high, cepstrum_formula)

  File d:\dev\ml_tools\ml_tools\audio\machine_learning\features.py:58 in compute
    self._computation()

  File d:\dev\ml_tools\ml_tools\audio\machine_learning\features.py:139 in _computation
    result = eval(feature["function"])(self.signals[domain]["y"], function_args)

  File d:\dev\ml_tools\ml_tools\audio\machine_learning\features_functions.py:1060 in ttn_ratio
    t_tnr, tnr, prom, tones_freqs = tnr_ecma_st(signal, args_dict["fs"], prominence=True)

  File ~\Anaconda3\lib\site-packages\mosqito\sq_metrics\tonality\tone_to_noise_ecma\tnr_ecma_st.py:39 in tnr_ecma_st
    tones_freqs, tnr, prom, t_tnr = _tnr_main_calc(spectrum_db, freq_axis)

  File ~\Anaconda3\lib\site-packages\mosqito\sq_metrics\tonality\tone_to_noise_ecma\_tnr_main_calc.py:76 in _tnr_main_calc
    peak_index = _screening_for_tones(freqs, spec_db, "smoothed", 90, 11200)

  File ~\Anaconda3\lib\site-packages\mosqito\sq_metrics\tonality\tone_to_noise_ecma\_screening_for_tones.py:55 in _screening_for_tones
    smooth_spec = _spectrum_smoothing(freqs, spec_db.T, 24, low_freq, high_freq, freqs).T

  File ~\Anaconda3\lib\site-packages\mosqito\sq_metrics\tonality\tone_to_noise_ecma\_spectrum_smoothing.py:65 in _spectrum_smoothing
    smoothed_spectrum = np.delete(smoothed_spectrum, i, axis=1)

  File <__array_function__ internals>:5 in delete

  File ~\Anaconda3\lib\site-packages\numpy\lib\function_base.py:4523 in delete
    raise IndexError(

IndexError: index 1 is out of bounds for axis 1 with size 1

After doing some probing in the scripts I noticed that the following (line 65 of _spectrum_smoothing) was the cause of the error:
smoothed_spectrum = np.delete(smoothed_spectrum, i, axis=1)

At the second band, it enters the if condition at line 64 and because the smoothed_spectrum array was instantiated with a size of (168, 1), it can't delete along axis=1.

Shouldn't it be axis=0? as is the case for the following delete in line 66?

It works when I pass longer snapshots (I tried 400ms and more) but mainly because it doesn't pass the if condition.

[wantysal]

Hi Antoine,

Thanks for this remark, your work is helping us with making mosqito more robust to all possible cases ;)
The correction has been included in PR#72 that will lead to the next coming release.

EDIT: merge done on 07/02/2024 including the correction