An Integrated, Cognitive RF Sensor Computational Model for Inferential Detection of Conversational Behavior

An integrated sensor framework is described and simulated which demonstrates an intelligent or cognitive RF sensor detection method that detects unique stochastic behaviors of incoming event data. For signal surveillance / RF spectrum management sensors, detecting interesting human behavior and rapidly adapting the sensor to optimally process these behaviors is the challenge. The detection method described integrates statistical and inference techniques in a novel way to detect the human factors of “who is talking to whom” in a field of seemingly random signaling channels. In this paper, we introduce CEMDIC (Correlation Enhanced Markov Detection and Inferential Convergence), a novel system construct of integrating correlation, hidden Markov models, and Bayesian inference sensor detection to rapidly detect conversational turn-taking and adapt sensor resources. Simulations and examples based upon radio frequency (RF) spectral data are used to illustrate and validate discussed methods and viability as a constrained sensor detection method.