pitch_shift.py

import pandas as pd
import os
import librosa
import librosa.display
import cv2
import numpy as np
import soundfile as sf

### Function for generatng pitch_shifted audio samples

def pitchShift(meta_data_path,audio_data_path,augmented_signals):
    metaDataPath = "/Users/alex/Desktop/coughvid/meta_data.csv"
    audioDataPath = "/Users/alex/Desktop/coughvid/data/audio/wavs-silence-removed/"
    augmentedSignals = "/Users/alex/Desktop/coughvid/data/audio/augmented/audio/"
    metaData = pd.read_csv(metaDataPath)
    counter = 0
    for index,row in metaData.iterrows():
        fname = row["uuid"]
        print(fname, " ", str(index+1),"/",str(metaData.shape[0]))
        signal , sr = librosa.load(audioDataPath+fname+".wav")
        ## Cough detection refinment: greater than 0.7
        if row["cough_detected"] >= 0.7:
            ## Multi-class to binary classification:
            if row["status"]=="COVID-19" or row["status"] == "symptomatic":
                sf.write(augmentedSignals+"sample{0}_{1}.wav".format(counter,1), signal, sr,'PCM_24')
                counter+=1
                pitch_shifting = librosa.effects.pitch_shift(signal,sr,n_steps=-4)
                sf.write(augmentedSignals+"sample{0}_{1}.wav".format(counter,1),pitch_shifting, sr,'PCM_24')
                counter+=1
            else:
                sf.write(augmentedSignals+"sample{0}_{1}.wav".format(counter,0), signal, sr,'PCM_24')
                counter+=1

metaDataPath = "/Users/alex/Desktop/coughvid/meta_data.csv"
audioDataPath = "/Users/alex/Desktop/coughvid/data/audio/wavs-silence-removed/"
augmentedSignals = "/Users/alex/Desktop/coughvid/data/audio/augmented/audio/"

pitchShift(metaDataPath,audioDataPath,augmentedSignals)

### Pitch-shift applied