MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641108185 A) filed by Pragati Engineering College on September 09, 2026, for Ai Flock Stress & Disease Early Warning For Poultry Sheds.
Inventors include Dr. A. Radha Krishna; Mrs. V Anantha Lakshmi; Mr. Bomma Bharath Kumar; and Dr. Alugolu Avinash.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: The present invention relates to an AI-powered early-warning system for detecting poultry flock disease-associated respiratory distress and heat stress using multi- modal sensing and edge-based decision fusion. A shed-mounted node comprises an INMP441 MEMS I2S microphone, an ESP32-S3 N16R8 controller having 8 MB PSRAM, an MQ-135 ammonia-related sensor, a DHT22 temperature and humidity sensor, a SIM800L GSM module, a MicroSD module, status LEDs and a regulated power supply with optional lithium-ion backup. The microphone acquires 16 kHz mono flock audio in approximately 2–4 second overlapping windows. MFCC features with delta and delta-delta coefficients are calculated locally and supplied to a compact depthwise-separable CNN implemented using TFLite-Micro. The classifier identifies normal, distress-respiratory, distress-heat and alarm-acute states. In parallel, environmental data are processed by a lightweight statistical detector using rolling z-score and slope over approximately 6–12 hour windows to identify NH3_rising_trend and heat_index_critical states. The principal technical feature is decision-level fusion with temporal persistence and hysteresis. A distress-respiratory state combined with a sustained NH3_rising_trend produces a Disease Onset Alert after prescribed persistence, whereas distress-heat combined with a critical heat index produces a Heat Stress Alert. An alarm-acute state without environmental corroboration is suppressed from disease SMS transmission and logged for later analysis. The system preferably performs inference locally, uses a quantized model targeted below 100 KB and below 50 ms inference time, stores operational data on MicroSD and transmits alerts by SMS without requiring a smartphone application or cloud service. The arrangement is intended to provide an early-warning target approximately 12–24 hours before visible mortality, subject to field validation, and is particularly suited to cost-sensitive poultry sheds, including environments using indigenous Indian breeds.
Disclaimer: Curated by HT Syndication.
