ABSTRACT
Hidden Markov Models (HMMs) are a common tool in pattern recognition. Applications of HMMs include voice recognition, texture recognition, handwriting recognition, gait recognition, tracking, and human behavior recognition. Variations of these applications can also be used in distributed sensor networks. The chapter introduces the methods of inferring HMMs from data streams. It provides background on HMMs and explains several HMM inference algorithms. In the applications, HMMs are inferred from data streams in sensor network using different approaches. K. L. Shalizi’s approach finds statistically significant groupings of the training data that correspond to HMM states. The drawback of Baum–Welch algorithm is that it requires a priori knowledge of the structure of the HMM. Shalizi developed the causal state splitting and reconstruction algorithm to infer HMMs without this prior information. HMMs have been used in sensor networks as well for a wide range of applications, that is, target action recognition, target tracking, sensor localization, and side channel analysis.
