MUMBAI, India, Oct. 12 -- Intellectual Property India has published a patent application (202641118419 A) filed by Koneru Lakshmaiah Education Foundation on October 03, 2026, for Noise-Robust Deep Learning Speech-To-Text Recognition System.
Inventors include Mounika Yalavarthi; Sunkara Srinidhi; Mohammad Ismail Zabeeullah; Geethana Yerramsetty; and Manoharmayum Dolpriya.
The application for the patent was published on October 09, 2026, under issue no. 41/2026.
Abstract: The present invention relates to a deep learning-based automatic speech recognition system for real-time conversion of speech into text. The system comprises a speech acquisition module, preprocessing module, acoustic feature extraction module, deep learning recognition module, decoding module, language processing module and text output module. An acquired speech signal is subjected to noise reduction, silence removal, amplitude normalization, segmentation and framing, followed by extraction of acoustic features including Mel-Frequency Cepstral Coefficients (MFCCs). The extracted features are processed using one or more deep learning architectures including CNN, LSTM and Transformer architectures to generate recognition outputs. The outputs are decoded and processed using linguistic information to generate corresponding text, thereby facilitating real-time speech recognition under varying acoustic conditions.
Disclaimer: Curated by HT Syndication.
