A structure-adaptive piece-wise linear segments representation for time series

This paper presents a structure-adaptive piece-wise linear segments representation of time series. The 1-th order landmarks are made as the endpoints of the liner segment by computing an error criterion, this algorithm can automatically produce the K piece-wise segments of time series, which can approximate the time series. This representation allows efficient computation of the similar measure. And we present a method of the similar measure, which is designed to be insensitive to noise, shifting, amplitude scaling and time scaling. The k-mean clustering algorithm is run on this representation. The results show that the representation can improve the clustering precision.