Time normalization of action time series using their wavelet coefficients-for generation of humanoid robots' actions with fertile emotions

Abstract

Time series is composed of many pairs of values and times. The time means when the value appears. Suppose that we observe same phenomena plural times and obtain plural time series. Although the same values happen plural times, their appearing times fluctuate. The problem to correct these time fluctuation is called "time normalization". We propose a new method of time normalization and apply it to action time series of a Bunraku puppet.