MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641074841 A) filed by Mlr Institute Of Technology on June 17, 2026, for System And Method For Speaker Identification Using Feature Fusion And Hybrid Deep Learning Architecture.
Inventors include Dr. Sivaramakrishna Yechuri; Dr. S V S Prasad; Ms. P. Vidya; and Mr. S. Sandeep.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: The present invention discloses a system and method for speaker identification using feature fusion and a hybrid deep learning architecture. The system extracts Mel Frequency Cepstral Coefficients (MFCC) and Mel spectrogram features from input speech signals and combines them to form a unified feature representation that captures both compact spectral and detailed time-frequency information. The fused features are processed using a hybrid deep learning model comprising convolutional layers, residual learning blocks, multi-head attention mechanisms, and bidirectional Long Short-Term Memory (Bi-LSTM) networks to effectively capture spatial, temporal, and contextual characteristics of speech. The proposed system enhances speaker discrimination, particularly for short-duration speech signals, and achieves high accuracy and robustness under varying acoustic conditions. The invention is suitable for real-time applications in biometric authentication, voice-based security systems, and intelligent human–machine interaction platforms.
Disclaimer: Curated by HT Syndication.
