MUMBAI, India, Jan. 9 -- Intellectual Property India has published a patent application (202541133078 A) filed by Dr. S. Maheswari; Dr. S. Sankar; Dr. C. Murugesan; Ms. Shobana D; Dr. V. Samuthira Pandi; Mr. Vadde Seetharama Rao; Ms. Jeevitha R; Mr. Aleemullakhan Pathan; Dr. Tephillah S; and Mr. Allan Dino J, Chennai, Tamil Nadu, on Dec. 29, 2025, for 'smart semiconductor sensors with embedded machine learning for real-time data analytics.'
Inventor(s) include Dr. S. Maheswari; Dr. S. Sankar; Dr. C. Murugesan; Ms. Shobana D; Dr. V. Samuthira Pandi; Mr. Vadde Seetharama Rao; Ms. Jeevitha R; Mr. Aleemullakhan Pathan; Dr. Tephillah S; and Mr. Allan Dino J.
The application for the patent was published on Jan. 9, under issue no. 02/2026.
According to the abstract released by the Intellectual Property India: "This invention is about smart semiconductor sensor devices and systems that have on-chip sensing, signal conditioning, and built-in machine learning processors. These systems allow for real-time analytics at or near the sensing interface. The invention solves problems with traditional sensor systems that use remote servers or cloud in-frastructure to process data. These systems have higher latency, use more bandwidth, are less private, and are not very good for Internet-of-Things (IoT) deployments where high-frequency data streams are always being sent. The invention features a heterogeneous integrated circuit architecture in which one or more semiconductor sensing elements (100) are monolithically or hybridly integrated with an analog front-end block (110), an analog-to-digital conversion block (120), a digital signal processing block (130), a non-volatile memory block (140), and a dedicated on-chip machine learning accelerator block (150). The accelerator block (150) is set up to run trained models like artificial neural networks, gradient-boosted trees, or other statis-tical learning models on streaming sensor data in real time to do things like feature extraction, classification, regression, anomaly detection, or predictive maintenance inference. The invention also includes a reconfigurable machine learning framework that lets you change the parameters of the embedded models either locally, using incremental learning mechanisms, or remotely, through a communication interface block (160) that receives secure over-the-air updates. A hierarchical data analytics pipeline is set up. It starts with low-level noise reduction, filtering, and normalization in the analog front-end block (110) and digital signal processing block (130). Then, it moves on to mid-level feature computation in the machine learning accelerator block (150) and high-level decision logic in the control and interface block (170). The smart semi-conductor sensor device (10) that comes out of this process can independently calcu-late actionable indicators, flags, and control signals. This cuts down on the amount of raw data sent to external systems and makes it possible to have deterministic local control loops. The invention can be used with many different types of sensors, such as temperature, pressure, strain, vibration, acoustic, optical, gas, bio-chemical, and multi-modal sensor arrays. In some cases, the smart semiconductor sensor device (10) is set up as part of a dis-tributed network of nodes. Each node does local analytics and sends summary statis-tics, model confidence metrics, and compressed features to a cloud-based or edge-based analytics framework on a regular basis. This kind of framework can manage model updates, fleet-level monitoring, and cross-device transfer learning to make performance better over time without needing to send raw sensor data all the time. The invention makes it possible to do real-time data analytics in industrial automa-tion, healthcare monitoring, smart energy, transportation, and consumer electronics applications that are scalable, have low latency, and protect privacy."
Disclaimer: Curated by HT Syndication.
