Footwear based wearable sensors are used in applications such as activity monitoring, gait analysis, post-stroke rehabilitation, body weight measurements, and energy expenditure estimation. Such wearable sensors typically required the modification or alteration of the shoe, which is not typically feasible for large populations without the direct involvement of shoe manufacturers. This dissertation presents an insole-based wearable sensor (SmartStep) that has its electronics fully embedded into a generic insole, which can be used with a large variety of shoes and, thus, resolves the need for shoe modification. The SmartStep is an always-on electronic device that comprises of inertial sensors and resistive pressure sensors implemented around a system on chip with Bluetooth low energy (BLE) connectivity. The dissertation explains a novel Android application methodology to interface multiple BLE sensors. As a part of the research, SmartStep’s predecessor, SmartShoe, was tested in an application scenario of gait monitoring of children with cerebral palsy. Novel signal processing algorithms were developed to deal with the effect of orthotics on pressure sensors. Machine learning models were used to recognize 3 major activities (sitting, standing and walking) with greater than 95% accuracy. The gait parameter estimation resulted in average error less than 6% across a range of temporal gait parameters. The SmartStep device was utilized in a study for recognizing activities of daily living (ADL) in adults. An upper-body, wrist-worn sensor was also utilized in the same study to compare the effectiveness of upper and lower body-worn sensors in recognizing ADL. The indirect calorimetry data from a portable metabolic system (Cosmed K4b), collected in the same study, were used to develop energy expenditure (EE) estimation models. Two different sets of steady state activity branched models were developed, along with models for transition activities. The EE models were validated on the data from a room calorimeter. SmartStep was 90% accurate in recognizing a broad set of nine ADLs, including household activities; while having 6% validation error in estimating total energy expenditure. These results suggest that SmartStep is accurate in recognizing multiple activities of daily living and in energy expenditure estimation.