A wavelet-based method for extracting intermittent discontinuities observed in human motor behavior. (February 2015)

• Record Type: Journal Article
• Authors: Inoue, Yasuyuki; Sakaguchi, Yutaka
• Is Part Of: Neural networks. Volume 62(2015:Feb.)
• Page Start: 91
• Page End: 101
• Publication Date: 2015-02
• Subjects: Motor intermittency -- Intermittent discontinuities -- Sub-movement -- Wavelet analysis -- Discontinuity detection; Neural computers -- Periodicals; Neural networks (Computer science) -- Periodicals; Neural networks (Neurobiology) -- Periodicals; Nervous System -- Periodicals; Ordinateurs neuronaux -- Périodiques; Réseaux neuronaux (Informatique) -- Périodiques; Réseaux neuronaux (Neurobiologie) -- Périodiques; Neural computers; Neural networks (Computer science); Neural networks (Neurobiology); Periodicals; 006.32
• DOI: 10.1016/j.neunet.2014.05.004
• Languages: English
• ISSNs: 0893-6080

Abstract

Abstract: Human motor behavior often shows intermittent discontinuities even when people try to follow a continuously moving target. Although most previous studies revealed common characteristics of this "motor intermittency" using frequency analysis, this technique is not always appropriate because the nature of the intermittency is not stationary, i.e., the temporal intervals between the discontinuities may vary irregularly. In the present paper, we propose a novel method for extracting intermittent discontinuities using a continuous wavelet transform (CWT). This method is equivalent to the detection of peak of the jerk profile in principle, but it successfully and stably detects discontinuities using the amplitude and phase information of the complex wavelet transform. More specifically, the singularity point on the time-scale plane plays a key role in detecting the discontinuities. Another important feature is that the proposed method does not require parameter tuning because it is based on the nature of hand movement. In addition, this method does not contain any optimization process, which avoids explosive increase in computational cost for long time-series data. The performance of the proposed method was examined using an artificial trajectory composed of several primitive movements, and an actual hand trajectory in a continuous target-tracking task. The functional rationale of the proposed method is discussed.