141 lines
4.8 KiB
Python
141 lines
4.8 KiB
Python
import glob
|
|
import numpy as np
|
|
import matplotlib.pyplot as plt
|
|
from thunderhopper.modeltools import load_data, save_data
|
|
from thunderhopper.filetools import crop_paths
|
|
from thunderhopper.filtertools import find_kern_specs
|
|
from thunderhopper.model import process_signal
|
|
from IPython import embed
|
|
|
|
# GENERAL SETTINGS:
|
|
target = 'Omocestus_rufipes'
|
|
data_paths = glob.glob(f'../data/processed/{target}*.npz')
|
|
stages = ['raw', 'filt', 'env', 'log', 'inv', 'conv', 'bi', 'feat']
|
|
save_path = '../data/inv/full/'
|
|
|
|
# ANALYSIS SETTINGS:
|
|
example_scales = np.array([0, 1, 10, 50])
|
|
scales = np.geomspace(0.01, 100, 100)
|
|
scales = np.unique(np.concatenate((scales, example_scales)))
|
|
kernels = np.array([
|
|
[1, 0.002],
|
|
[-1, 0.002],
|
|
[2, 0.004],
|
|
[-2, 0.004],
|
|
[3, 0.032],
|
|
[-3, 0.032]
|
|
])
|
|
kernels = None
|
|
types = None#np.array([-1])
|
|
sigmas = None#np.array([0.001, 0.002, 0.004, 0.008, 0.016, 0.032])
|
|
|
|
# EXECUTION:
|
|
for data_path, name in zip(data_paths, crop_paths(data_paths)):
|
|
print(f'Processing {name}')
|
|
|
|
# Get song recording:
|
|
data, config = load_data(data_path, files='raw')
|
|
song, rate = data['raw'], config['rate']
|
|
|
|
# Reduce to kernel subset:
|
|
kern_inds = find_kern_specs(config['k_specs'], kernels, types, sigmas)
|
|
config['kernels'] = config['kernels'][:, kern_inds]
|
|
config['k_specs'] = config['k_specs'][kern_inds, :]
|
|
config['k_props'] = [config['k_props'][i] for i in kern_inds]
|
|
config['feat_thresh'] = config['feat_thresh'][kern_inds]
|
|
|
|
# Get song segment to be analyzed:
|
|
time = np.arange(song.shape[0]) / rate
|
|
start, end = data['songs_0'].ravel()
|
|
segment = (time >= start) & (time <= end)
|
|
|
|
# Normalize song component:
|
|
song /= song[segment].std(axis=0)
|
|
|
|
# Get normalized noise:
|
|
rng = np.random.default_rng()
|
|
noise = rng.normal(size=song.shape[0])
|
|
noise /= noise[segment].std()
|
|
|
|
# Prepare snippet storage:
|
|
shape_low = (song.shape[0], example_scales.size)
|
|
shape_high = (song.shape[0], config['k_specs'].shape[0], example_scales.size)
|
|
snippets = dict(
|
|
snip_raw=np.zeros(shape_low, dtype=float),
|
|
snip_filt=np.zeros(shape_low, dtype=float),
|
|
snip_env=np.zeros(shape_low, dtype=float),
|
|
snip_log=np.zeros(shape_low, dtype=float),
|
|
snip_inv=np.zeros(shape_low, dtype=float),
|
|
snip_conv=np.zeros(shape_high, dtype=float),
|
|
snip_bi=np.zeros(shape_high, dtype=float),
|
|
snip_feat=np.zeros(shape_high, dtype=float)
|
|
)
|
|
|
|
# Prepare measure storage:
|
|
shape_low = (scales.size,)
|
|
shape_high = (scales.size, config['k_specs'].shape[0])
|
|
measures = dict(
|
|
measure_raw=np.zeros(shape_low, dtype=float),
|
|
measure_filt=np.zeros(shape_low, dtype=float),
|
|
measure_env=np.zeros(shape_low, dtype=float),
|
|
measure_log=np.zeros(shape_low, dtype=float),
|
|
measure_inv=np.zeros(shape_low, dtype=float),
|
|
measure_conv=np.zeros(shape_high, dtype=float),
|
|
measure_feat=np.zeros(shape_high, dtype=float)
|
|
)
|
|
|
|
# Execute piecewise:
|
|
for i, scale in enumerate(scales):
|
|
print('Simulating scale ', scale)
|
|
|
|
# Rescale song and add noise:
|
|
scaled = song * scale + noise
|
|
|
|
# Process mixture:
|
|
signals, rates = process_signal(config, returns=stages,
|
|
signal=scaled, rate=rate)
|
|
# Store results:
|
|
for stage in stages:
|
|
key = f'measure_{stage}'
|
|
|
|
# Log snippet data:
|
|
if scale in example_scales:
|
|
scale_ind = np.nonzero(example_scales == scale)[0][0]
|
|
snippets[f'snip_{stage}'][:, ..., scale_ind] = signals[stage]
|
|
|
|
# Log intensity measure per stage (excluding binary):
|
|
if stage in ['raw', 'filt', 'env', 'log', 'inv', 'conv']:
|
|
measures[key][i] = signals[stage][segment, ...].std(axis=0)
|
|
elif stage == 'feat':
|
|
measures[key][i] = signals[stage][segment, :].mean(axis=0)
|
|
|
|
# thresh_y = np.percentile(measures['measure_feat'], 99, axis=0)
|
|
# kern_types = np.unique()
|
|
# thresh_x = np.zeros(thresh_y.shape, dtype=float)
|
|
# for i, thresh in enumerate(thresh_y):
|
|
# if thresh < 0.1:
|
|
# thresh_x[i] = scales[-1]
|
|
# continue
|
|
# mask = (measures['measure_feat'][:, i] < thresh)
|
|
# thresh_x[i] = scales[np.nonzero(mask)[0][-1]]
|
|
# inds = np.argsort(thresh_x)
|
|
# print(config['k_specs'][inds, :])
|
|
|
|
# fig, axes = plt.subplots(1, 2)
|
|
# axes[0].plot(snippets['snip_feat'][:, inds, -1])
|
|
# axes[1].plot(scales, measures['measure_feat'][:, inds])
|
|
# plt.show()
|
|
# embed()
|
|
|
|
# Save analysis results:
|
|
if save_path is not None:
|
|
data = dict(
|
|
scales=scales,
|
|
example_scales=example_scales,
|
|
)
|
|
data.update(snippets)
|
|
data.update(measures)
|
|
save_data(save_path + name, data, config, overwrite=True)
|
|
print('Done.')
|
|
embed()
|