Design Exploration of a Human-machine Interface Application

A data glove is a human-machine interface designed to capture the motion of a computer operator's hand for real- time input. Current implementations of such devices are commonly bulky, wired, and have high power requirements. It has been proposed that a data glove could be implemented with low power, integrated sensing and processing components with wireless communications. In this paper, we evaluate an existing lower-power data-glove implementation. Then, with minimizing power requirements as a goal, we evaluate design alternatives that tradeoff application specific local processing with communication. Finally, we compare the performance of those software alternatives on several general purpose microcontrollers and DSPs. current. At that size, the motes could be placed on the tips of fingers and on the back of the hand for a much less obtrusive data glove interface. In this paper, we first evaluate the design and power requirements of the proof-of-concept prototype glove developed by the SmartDust group, called the Acceleration Sensing Glove (ASG). Then, based on the results of the analysis, we define an exploration methodology for evaluating alternative implementations of the ASG. In particular, since SmartDust motes have processors integrated with the sensors, we investigate the tradeoff between local computation and communication. We perform this evaluation in the context of three different processor architectures: the Atmel AVR 8-bit RISC microcontroller (2), the Texas Instruments MSP430 16-bit microcontroller (9), and the Texas Instruments TMS320C5000 DSP platform (12). Finally, we discuss the results of the analysis, some directions for future work, and conclusions.