MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641111354 A) filed by Koneru Lakshmaiah Education Foundation on September 17, 2026, for Adaptive Pooling-Based Speech And Music Classification System.
Inventors include Ch. V. Sai Abhiram; N. Vihas; G. Ruthika Lakshmi; D. Udayanjali; and Akhilesh Kumar Dubey.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: A method and system are disclosed for segment-level classification of audio recordings as speech or music using Wav2Vec 2.0 acoustic representations and a Bidirectional Long Short-Term Memory (BiLSTM) network. An input audio waveform is processed to obtain contextual acoustic representations, which are grouped into temporal segments and processed by the BiLSTM network to generate hidden-state vectors. A plurality of candidate pooled representations are generated using attention, mean, maximum, and last-hidden-state pooling. An Automatic Pooling Selection module independently evaluates the candidate pooling strategies based on classification performance and selects the highest-performing strategy. The selected representation is supplied to a classification network to generate a speech/music decision output.
Disclaimer: Curated by HT Syndication.
