MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202611117031 A) filed by Think Mod Ai Pvt. Ltd. on September 30, 2026, for Multi-Stage Cross-Attention System For Adaptive Processing Of Phonocardiogram Signals.
Inventors include Pranshu Tiwari; Vanshika Mehlawat; Mrinal Bamhotra; and Divyashree Naveen.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: An adaptive multi-stage phonocardiogram (PCG) signal processing system and method for selective noise suppression are disclosed. An input interface receives a PCG signal, which is transformed into a time-frequency representation using a Constant-Q Transform (CQT). A first U-Net denoising stage employs Multi-Stage Frequency- Time Attention (MSFTA) to generate a first denoised representation. A mid-stage nodal classifier generates a probability value indicative of a candidate abnormal cardiac acoustic component. Based on the probability value, an adaptive gating module selectively applies or bypasses Shannon-entropy noise reduction. A second U- Net denoising stage applies Feature-Wise Linear Modulation (FiLM) parameters generated from a conditioning representation to generate a second denoised representation. A confidence-fusion module dynamically combines the denoised representations according to signal-quality-dependent confidence values. An inverse CQT generates a reconstructed time-domain PCG signal having reduced noise while retaining cardiac acoustic components.
Disclaimer: Curated by HT Syndication.
