Answers are grounded in this paper and cite their sources.
New segmentation technology enhances accuracy in detecting speech segments amidst background noise, indicating practical applications in voice interfaces.
«Signal/pause» segmentation is a key task in speech signal processing, which consists in determining the exact boundaries between speech and pauses. The influence of such a factor as background noise significantly complicates this process, since it can distort the true boundaries of speech segments and pauses. It is necessary to develop a reliable segmentation technology that ensures high reliability of speech segment detection in the presence of background noise. Work purpose – development and study of «signal/pause» segmentation technology that allows for effective differentiation of the levels of mixing of speech signal fragments and reliable detection of the boundaries of speech segments and pauses. The study of the technology yielded results demonstrating a high level of reliability in determining the boundaries of speech and pauses. The best segmentation results with errors of 1.8% and 0.9% are achieved when compared with the values of the average level of fragment mixing and the median of the first 20 fragments corresponding to the initial pause with background noise. The proposed «signal/pause» segmentation technology has sufficient practical value, since its application allows for a significant increase in real-time reliability and a decrease in computational load. This is especially important for speech applications that provide human-computer interactions via voice interfaces.
No takes yet. Share an insight, caveat, or question.
Alimuradov et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: