In the analysis process of popular music singing audio, factors such as environmental noise interference and complex instrument accompaniment seriously affect the accuracy of audio feature extraction, resulting in the performance of traditional music beat extraction methods being difficult to meet practical needs.Therefore, this study innovatively proposes a popular music singing beat extraction method based on multifeature fusion.Performing pre-processing operations such as discretisation, denoising and normalisation on the original singing audio signal effectively improves signal quality.Through joint time-frequency domain analysis, comprehensively extract the time-frequency characteristics of music signals.Adopting a feature fusion strategy, combined with beat cycle analysis and inter beat distance calculation, highprecision beat detection is achieved.Experimental data shows that the missed detection rate and false detection rate of this method are as low as 2.1% and 2.5%, respectively, significantly better than traditional methods, providing reliable technical support for pop music performance analysis.
Kong et al. (Thu,) studied this question.