Optimized speaker change detection approach for speaker segmentation towards speaker diarization based on deep learning

A Fractional Ebola Optimization Search Algorithm Approach for Enhanced Speaker Diarization

“…The pertinent studies are summarized as follows: VijayKumar and Rao [3] developed a bottom-up speaker clustering technique enabled by active learning, aiming to improve diarization performance with minimal human intervention. Their approach comprised two phasesexploratory clustering and constrained clustering-wherein an active learning algorithm was utilized to generate precise speaker models and facilitate audio grouping.…”

“…The following encapsulates the contributions and shortcomings of each study: ➢ In the study by VijayKumar and Rao [3], an expected speaker-error-based segment selection approach was employed, demonstrating superior efficacy over random segment selection. However, this strategy failed to accommodate the study of human errors, leading to potential shortcomings in system performance.…”

“…An audio file comprises various sources, requiring an indexing process that categorizes the signal into speech and non-speech signals, the latter of which includes silence, music, and noise [3]. With the exponential growth of recorded speech, encompassing audio broadcasts, voice messages, and television, speaker diarization has emerged as a significant task.…”

“…With the exponential growth of recorded speech, encompassing audio broadcasts, voice messages, and television, speaker diarization has emerged as a significant task. The objective of this technique is to partition the speech signal and isolate signals from the same speaker, essentially answering "who spoke what and when" [3].…”

“…Conversely, top-down clustering begins by treating the entire audio as a single entity and progressively divides it into subclusters. Both strategies, while effective, are continuous processes and face the limitation of error propagation [3].…”