Homec4science
Diffusion COUGHVID (master) Browse

R10770/src/segmentation.py
1517d05b5315master

Last Change
Raw File

segmentation.py
View Options

import numpy as np
#Use old segmentation
def segment_cough(x,fs, cough_padding=0.2,min_cough_len=0.2, th_l_multiplier = 0.1, th_h_multiplier = 2):
"""Preprocess the data by segmenting each file into individual coughs using a hysteresis comparator on the signal power
Inputs:
*x (np.array): cough signal
*fs (float): sampling frequency in Hz
*cough_padding (float): number of seconds added to the beginning and end of each detected cough to make sure coughs are not cut short
*min_cough_length (float): length of the minimum possible segment that can be considered a cough
*th_l_multiplier (float): multiplier of the RMS energy used as a lower threshold of the hysteresis comparator
*th_h_multiplier (float): multiplier of the RMS energy used as a high threshold of the hysteresis comparator
Outputs:
*coughSegments (np.array of np.arrays): a list of cough signal arrays corresponding to each cough
cough_mask (np.array): an array of booleans that are True at the indices where a cough is in progress"""
cough_mask = np.array([False]*len(x))
#Define hysteresis thresholds
rms = np.sqrt(np.mean(np.square(x)))
seg_th_l = th_l_multiplier * rms
seg_th_h = th_h_multiplier*rms
#Segment coughs
coughSegments = []
padding = round(fs*cough_padding)
min_cough_samples = round(fs*min_cough_len)
cough_start = 0
cough_end = 0
cough_in_progress = False
tolerance = round(0.01*fs)
below_th_counter = 0
for i, sample in enumerate(x**2):
if cough_in_progress:
if sample<seg_th_l:
below_th_counter += 1
if below_th_counter > tolerance:
cough_end = i+padding if (i+padding < len(x)) else len(x)-1
cough_in_progress = False
if (cough_end+1-cough_start-2*padding>min_cough_samples):
coughSegments.append(x[cough_start:cough_end+1])
cough_mask[cough_start:cough_end+1] = True
elif i == (len(x)-1):
cough_end=i
cough_in_progress = False
if (cough_end+1-cough_start-2*padding>min_cough_samples):
coughSegments.append(x[cough_start:cough_end+1])
else:
below_th_counter = 0
else:
if sample>seg_th_h:
cough_start = i-padding if (i-padding >=0) else 0
cough_in_progress = True
return coughSegments, cough_mask
def compute_SNR(x, fs):
"""Compute the Signal-to-Noise ratio of the audio signal x (np.array) with sampling frequency fs (float)"""
segments, cough_mask = segment_cough(x,fs)
RMS_signal = 0 if len(x[cough_mask])==0 else np.sqrt(np.mean(np.square(x[cough_mask])))
RMS_noise = np.sqrt(np.mean(np.square(x[~cough_mask])))
SNR = 0 if (RMS_signal==0 or np.isnan(RMS_noise)) else 20*np.log10(RMS_signal/RMS_noise)
return SNR
c4science · Help