Multi-scale analysis of myoelectric signals

The neuromuscular system is composed of multiple units with highly non-linear deterministic and random characteristics. Such systems generate complex signals that exhibit different characteristics when they are analyzed on different scales, known as multi-scale signals. The myoelectric signal, which is comprised of deterministic (i.e., structured) and random contributions, exhibits multi-scale characteristics. In this work, we present four major paradigms for multi-scale analysis of myoelectric signals; namely, self-similarity, long memory, multi-fractality and chaos.;Long memory. A method for quantitative analysis of long memory, known as detrended fluctuation analysis, is introduced in the context of myoelectric signals. We show the myoelectric signals show both mono- and multi-fractality. A new approach is presented to compute an optimum Hurst exponent for fatigue estimation.;Multi-fractality. A unique multi-fractal process known as multiplicative cascade multi-fractal is introduced in the context of myoelectric signals. A framework for analysis of myoelectric signals using this multi-fractal process is presented for discerning neuropathic conditions.;Chaos. An interesting framework known as power-law sensitivity to initial condition is used to analyze myoelectric signals. This new framework suggests the myoelectric signals are not fully random or fully chaotic but they resemble random fractals and chaotic motions on different scales. The Lempel-Ziv measure is introduced for quantitative analysis of deterministic complexity. We show the binary Lempel-Ziv measure might be affected by a false sense of complexity due to inadequacy of two symbols to characterize motor unit action potentials. A new ternary Lempel-Ziv measure is introduced which resolves limitations of the binary Lempel-Ziv measure. The Lempel-Ziv measure also provides unique characteristics well-suited for fatigue estimation.;Self-similarity. A novel multi-scale bi-phase power-law is introduced which accurately characterizes surface myoelectric signal power spectrum recorded during moderate contractions. We present a novel methodology, termed the bi-phase power spectrum method which provides unique parameters that are distinctly sensitive to force, joint angle, and fatigue. These parameters could be used as complementary information for conventional myoelectric parameters that are confounded during dynamic contractions.;We integrate these four aspects along with chaos and random fractal theory and provide a unified framework for multi-scale analysis of myoelectric signals.