Evaluation of muscular fatigue during bicycle ergometer exercise using the proportion time-series of principal components

We tried to evaluate a myoelectric threshold by increasing the work load during bicycle ergometer exercise. First of all, multichannel recordings allowed us to avoid the influences of innervation zones on myoelectric signals during dynamic exercise. Then, we acquired the mean power frequency (MPF), the average rectified value (ARV), and the conduction velocity (CV) of myoelectric signals and the heart rate (HR). Based on well-conditioned measurement, we estimated the time-series of principal components from multivariate parameters composed of the ARV, MPF, and CV. As a result, the proportion of the first principal component decreased with the work load during the latter trial in which muscle fatigue would be expected, whereas it remained high during the first trial