speaker diarization for telephone conversations using Resemblyzer

Question

I have audio recordings of telephone conversations, I used Resemblyzer it clusters audio based on speakers. the output is labelling, which is basically a dictionary of which person spoke when (speaker_label, start_time, end_time)

I need to segments audio out speaker wise based on the time in labelling. I've been working on this for a week.

from resemblyzer import preprocess_wav, VoiceEncoder
from pathlib import Path
import pickle
import scipy.io.wavfile
from spectralcluster import SpectralClusterer

audio_file_path = 'C:/Users/...'
wav_fpath = Path(audio_file_path)

wav = preprocess_wav(wav_fpath)
encoder = VoiceEncoder("cpu")
_, cont_embeds, wav_splits = encoder.embed_utterance(wav, return_partials=True, rate=16)
print(cont_embeds.shape)            

clusterer = SpectralClusterer(
    min_clusters=2,
    max_clusters=100,
    p_percentile=0.90,
    gaussian_blur_sigma=1)

labels = clusterer.predict(cont_embeds)

def create_labelling(labels, wav_splits):
    from resemblyzer.audio import sampling_rate
    times = [((s.start + s.stop) / 2) / sampling_rate for s in wav_splits]
    labelling = []
    start_time = 0

    for i, time in enumerate(times):
        if i > 0 and labels[i] != labels[i - 1]:
            temp = [str(labels[i - 1]), start_time, time]
            labelling.append(tuple(temp))
            start_time = time
        if i == len(times) - 1:
            temp = [str(labels[i]), start_time, time]
            labelling.append(tuple(temp))

    return labelling

labelling = create_labelling(labels, wav_splits)

Answer 1

this code helps a lot: first add a time_stamps.txt file containg time stamps to trim the audio on(time_stamps.txt file should be comma separated). then add the audio file name and it's format and it does the job. I find this on github, https://github.com/raotnameh/Trim_audio

import numpy as np
    from pydub import AudioSegment
    
    
    def trim(start,end,file_name,format_,i):
        t1 = start
        t2 = end
        t1 = t1 * 1000 #Works in milliseconds
        t2 = t2 * 1000
        newAudio = AudioSegment.from_wav(file_name + "." +format_)
        newAudio = newAudio[t1:t2]
        newAudio.export(file_name+ "_" + str(i) + '.wav', format=format_) #Exports to a wav file in the current path.
    
    if __name__ == '__main__':
    
        with open("time_stamps.txt", "rb") as file:
            contents = list(map(float,file.read().decode("utf-8").split(',').strip()))
    
        file_name = "male"
        format_ = "wav"
        for i in range(len(contents)):
            try :trim(contents[i],contents[i+1],file_name,format_,i)
            except : pass