MUMBAI, India, Jan. 23 -- Intellectual Property India has published a patent application (202511129047 A) filed by Vivek Ranjan; and Dayashankar Singh, Gorakhpur, Uttar Pradesh, on Dec. 19, 2025, for 'system and method for audio authenticity verification using spatio-temporal analysis.'
Inventor(s) include Vivek Ranjan; and Dayashankar Singh.
The application for the patent was published on Jan. 23, under issue no. 04/2026.
According to the abstract released by the Intellectual Property India: "The present disclosure relates to a system and method for detecting deepfake audio using GAN-based spatial-temporal feature learning to enhance security of digital communication infrastructure. The system receives an input audio signal and performs adaptive signal conditioning using Wiener filtering to suppress noise while preserving synthesis-specific artifacts. Multi-domain speech forensic features including spectral, pitch irregularity, source excitation, physiological glottal flow, formant, energy, loudness, and modulation spectrum features are extracted and synchronized. The extracted features are processed through a one-dimensional dilated spatial-temporal learning network comprising convolutional layers, spatial and temporal attention mechanisms, and bidirectional gated recurrent units to capture localized and long-range inconsistencies in speech signals. A Wasserstein generative adversarial network discriminator is employed to stably model distributional divergence between genuine and synthetic speech and to classify the audio signal as real or deepfake. The system further generates a trust-aware output with a confidence metric, enabling robust, scalable, and infrastructure-ready deepfake audio detection under real-world noise and replay conditions."
Disclaimer: Curated by HT Syndication.
