diff --git a/BBPs/2-manual_verification.py b/BBPs/2-manual_verification.py
index 54134273ea6294bc27bc2dc8f24b9fa6a160f6f9..2c3da3c932e990c7b4ada38f68469ddf6e92a3a5 100644
--- a/BBPs/2-manual_verification.py
+++ b/BBPs/2-manual_verification.py
@@ -109,8 +109,8 @@ for file in range(len(audio_paths)):
width=[0,max_length])[0]
Magnitude_audio = stft(signal_high, n_fft=nfft, hop_length=hop_length)
- #spectre = amplitude_to_db(np.abs(Magnitude_audio))
- spectre = pcen(np.abs(Magnitude_audio) * (2**31))
+ #spectrum = amplitude_to_db(np.abs(Magnitude_audio))
+ spectrum = pcen(np.abs(Magnitude_audio) * (2**31))
# show ICI of each BBP
n_BBP = np.unique(BBP_all[use,-1])
@@ -143,7 +143,7 @@ for file in range(len(audio_paths)):
axs[0].axis(xmin=lower, xmax=upper)
axs[0].tick_params('x', labelbottom=False, bottom=False)
- axs[1].imshow(spectre[::-1][:,int(lower/hop_length):int(upper/hop_length)],
+ axs[1].imshow(spectrum[::-1][:,int(lower/hop_length):int(upper/hop_length)],
aspect='auto', interpolation='none', cmap='jet',
extent=(0,(upper-lower)/sr,0,int(sr/2)))
axs[1].set_title("Spectrogram (dB scale)")
@@ -195,7 +195,7 @@ for file in range(len(audio_paths)):
plt.close('all')
del signal, signal_high, signal_peaks, tk_signal,\
- Magnitude_audio, spectre
+ Magnitude_audio, spectrum
dict_annot[audio_paths[file]] = BBP_manual
# Temp save (in case of a crash)
diff --git a/Whistles/1-Identification_of_whistles.py b/Whistles/1-Identification_of_whistles.py
index ec49c2dd6764976171e178859ea248bbe664ff9e..906ed0f605c6cff15c5025b70b7ea0dd93974904 100644
--- a/Whistles/1-Identification_of_whistles.py
+++ b/Whistles/1-Identification_of_whistles.py
@@ -71,14 +71,14 @@ for file in range(len(audio_paths)):
# resample
signal_dec = resample(signal, int(((stop-start)*new_sr)))
- # extract spectral informations
+ # extract spectrum informations
Magnitude_audio = stft(signal_dec, n_fft=nfft, hop_length=hop_length)
- spectre = np.copy(np.abs(Magnitude_audio[f_min:,:]))
- # PCEN could replace spectrogram in very noisy recordings
- #spectre_pcen = pcen(np.abs(Magnitude_audio) * (2**31), bias=10)[f_min:,:]
+ spectrum = np.copy(np.abs(Magnitude_audio[f_min:,:]))
+ # PCEN could replace spectrum in very noisy recordings
+ #spectrum_pcen = pcen(np.abs(Magnitude_audio) * (2**31), bias=10)[f_min:,:]
# Selection algorithm
- max_loc_per_bin_check1 = get_local_maxima(spectre, spectre, nrg_rap)[1]
+ max_loc_per_bin_check1 = get_local_maxima(spectrum, spectrum, nrg_rap)[1]
trajectories = get_trajectories(max_loc_per_bin_check1, dist_f=dist_f, dist_t=dist_t)
final_traj = select_trajectories(trajectories, taille_traj_min, min_acce, max_acce, verbose=0)
corrected_traj = sparsity_ridoff(final_traj, error_thresh=sparsity)
diff --git a/Whistles/WhistleUtils.py b/Whistles/WhistleUtils.py
index 33132e89f39d61b2b3f5a020a57d61ef72f2f03d..56b8271a34271d9fe9b223ae8c4ef017c5d893ba 100755
--- a/Whistles/WhistleUtils.py
+++ b/Whistles/WhistleUtils.py
@@ -112,11 +112,11 @@ def get_csv(csv_folder, slash="\\"):
return data_frame, sorted_names
#%% Trajectories algorithms
-def get_local_maxima(spectrogram, spectrogram2, hardness, threshold=10e-5):
+def get_local_maxima(spectrum, spectrum2, hardness, threshold=10e-5):
"""
Parameters
----------
- spectrogram : NUMPY ARRAY
+ spectrum : NUMPY ARRAY
Spectrogram (float values) of an audio signal.
hardness : INT or FLOAT
Number of times a value has to be above the geometric mean in order to be kept.
@@ -128,21 +128,21 @@ def get_local_maxima(spectrogram, spectrogram2, hardness, threshold=10e-5):
local_max2 : NUMPY ARRAY
Spectrogram with 1 & 0 indicating local maxima above hardness*geom_mean
"""
- local_max1 = np.zeros(spectrogram.shape, dtype=int)
- local_max2 = np.zeros(spectrogram.shape, dtype=int)
- geom_mean = gmean(gmean(spectrogram, axis=1))
- geom_mean0 = gmean(spectrogram, axis=0)
- geom_mean1 = gmean(spectrogram, axis=1)
- # geom_mean0, geom_mean1 = gmean(spectrogram), gmean(spectrogram)
-
- for each_bin in range(spectrogram.shape[1]):
- for freq in range(1, spectrogram.shape[0]-1):
- if (spectrogram2[freq, each_bin] > spectrogram2[freq-1, each_bin]) and \
- (spectrogram2[freq, each_bin] > spectrogram2[freq+1, each_bin]):
+ local_max1 = np.zeros(spectrum.shape, dtype=int)
+ local_max2 = np.zeros(spectrum.shape, dtype=int)
+ geom_mean = gmean(gmean(spectrum, axis=1))
+ geom_mean0 = gmean(spectrum, axis=0)
+ geom_mean1 = gmean(spectrum, axis=1)
+ # geom_mean0, geom_mean1 = gmean(spectrum), gmean(spectrum)
+
+ for each_bin in range(spectrum.shape[1]):
+ for freq in range(1, spectrum.shape[0]-1):
+ if (spectrum1[freq, each_bin] > spectrum2[freq-1, each_bin]) and \
+ (spectrum2[freq, each_bin] > spectrum2[freq+1, each_bin]):
local_max1[freq, each_bin] = 1
- if (spectrogram[freq, each_bin] > (threshold)):
- if (spectrogram[freq, each_bin] > (geom_mean0[each_bin]*hardness)):
- if (spectrogram[freq, each_bin] > (geom_mean1[freq]*hardness)):
+ if (spectrum[freq, each_bin] > (threshold)):
+ if (spectrum[freq, each_bin] > (geom_mean0[each_bin]*hardness)):
+ if (spectrum[freq, each_bin] > (geom_mean1[freq]*hardness)):
local_max2[freq, each_bin] = 1
return local_max1, local_max2