A novel framework that includes the time variant Hidden Markov Model. It estimates the user state using the concept of Human Activity Recognition to implement the context aware application. That application requires the continuous data acquisition and interpretation from sensor reading .Determine the accurate user state representation model and model operates with maximum power efficiency. Continuous running of sensor need to extend the lifetime of battery. The final framework is to achieve a fine balance in the defined trade-off analysis Depending on user profiles power accuracy range should be increase. The proposed framework is validated and implemented in a HAR-based application. The proposed framework shows an increase in power efficiency of 60% for an accuracy range from 75% upto 96%, depending on user profiles.