Adaptive Suppression of Tremor for Improved Human-machine Control

Cameron N. Riviere
Ph. D. dissertation, Department of Mechanical Engineering, The Johns Hopkins University, 1995.

Abstract

Tremor is involuntary, roughly sinusoidal movement. Physiological tremor causes imprecision in fine manual tasks such as microsurgery, limiting the quality of medical care. Pathological tremor afflicts hundreds of thousands of Americans due to disease and injury. Severe cases are completely debilitating. Effective tremor suppression is sorely needed to improve the quality of life of persons with pathological tremor, and to improve precision during microsurgery, resulting in less tissue damage, faster recovery, and lower health care costs.

The Fourier Linear Combiner (FLC) is an adaptive algorithm for estimating periodic signals of known frequency. The FLC forms a truncated Fourier series model of an incoming signal which can be used to cancel periodic interference. I have extended the FLC to the case of unknown fundamental frequency by developing the Weighted-frequency Fourier Linear Combiner (WFLC). In the WFLC, the reference signal frequency itself is adapted using a modification of the LMS algorithm.

I provide an analytical description of WFLC performance. The analysis focuses on the reference frequency weight, the quality of its estimate of primary signal frequency, and the effect of its accuracy on the capacity of the amplitude weights to model a periodic signal.

I present the application of the WFLC to modeling and canceling of tremor during human-machine control. It adapts to the unknown frequency of tremor and tracks frequency and amplitude modulation. The WFLC forms the basis for three practical applications. The first is adaptive canceling of pathological tremor for computer input using interfaces such as pens and sensory gloves. I also implement the WFLC in a desktop system which quantifies tremor for clinical and diagnostic use. Finally, I demonstrate active real-time compensation of physiological tremor for use in a handheld microsurgical instrument.

The WFLC effectively suppresses tremor, requiring no a priori assumptions about tremor frequency. It is zero-phase, and has an infinite null. Its computational simplicity makes it attractive for rehabilitative applications. Its explicit zero-phase model of tremor makes it useful for active tremor control, e.g., during microsurgery.


Carnegie Mellon Computer Science Cam.Riviere@cmu.edu