Micro-harvesting from sources such as indoor light can enable a plethora of self-sustainable sensing systems for mobile healthcare applications. However, given the minuscule and variable amount of energy harvested from these renewable sources, practical sensing systems powered by micro-harvesting is today limited to light driven motion sensing. In this paper, we design, implement, and evaluate an indoor light driven wearable glove device that uses flex sensors and accelerometers for hand gesture recognition. Through the design, we make a two-fold contribution to micro-harvester driven mobile sensing systems. First, motivated by extensive profiling of panels for indoor light scavenging, we design a harvester that multiplexes panels of different compositions to maximally scavenge energy as a function of lighting conditions. Second, we present a tiered architecture composed of application specific hardware logic, wakeup controllers, a general purpose micro-controller, and a bluetooth device that can adapt to variable and ultra-low energy constraints, and at the same time provide high responsiveness and compute capability for gesture recognition. We evaluate the glove device in the context of a hand gesture driven home automation system for the elderly.