A First-Order Markov Model for Wellness Mobile Applications

The concept of wellness mobile by incorporating continuous monitoring of biometrics on wireless handheld devices has been attracting a lot of attention from both industrialists and academia. In this paper, we propose a probabilistic inference algorithm for the wellness mobile applications by monitoring fluctuations in biometric(s) in order to provide effective inference about human wellness in an efficient and timely manner. Wellness state recognition can be achieved through dynamic probabilistic inference from the sensory data using multiple-modality sensors. Since the overriding goal is to embed this feature into a cellular phone, the proposed algorithm cannot be computationally complex and rely on too many biometrics. Here, the proposed algorithm relies on a single biometric, in which the dynamic probabilistic inference depends on a temporal window of observations of the biometric. Specifically, we monitor skin temperature variations during different activities in order to track the variations in human stress. The algorithm can be generalized by a temporal Bayesian network (TBN) framework. The simplicity of the technique makes it a favorable candidate for implementation on cellular phones with existing biometric sensors that are available in most smart phones today.