MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202621096620 A) filed by Bobhate, Grishma; and Bhaladhare, Pawan on August 10, 2026, for A Leakage-Aware Multimodal Behavioral Analytics Framework For Depression Screening And Mood Monitoring In The Internet Of Behavior.
Inventors include Bobhate, Grishma; and Bhaladhare, Pawan.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: The present invention is related to a leakage-aware multimodal behavioral analytics framework for depression screening and mood monitoring in the internet of behavior. The framework performs participant-level depression classification and mood-change prediction using participant-only interview data. Interviewer turns are removed from transcripts, and participant-level grouping prevents subject leakage across training and validation. The framework combines transcript embeddings generated from MiniLM, Sentence-BERT, or DistilBERT representations with Wav2Vec2-derived speech embeddings and COVAREP acoustic descriptors. Optional CLNF/OpenFace visual descriptors are reduced through principal component analysis before fusion. Feature-level concatenation supports calibrated LinearSVC classification and focal-loss-based neural fusion for threshold-aware depression screening. A Mood Signal Index (MSI) divides transcripts into chronological bins and derives mood-related sequence statistics including mean mood level, volatility, slope, negative-bin ratio, and final-minus-initial mood change. A GRU regressor predicts temporal mood change from MSI sequences for interpretable behavioral monitoring within IoB-based digital mentalhealth analytics.
Disclaimer: Curated by HT Syndication.
