Sensor networks have increased the amount and variety of temporal data available, requiring the definition of new techniques for data mining. Related research typically ad- dresses the problems of indexing, clustering, classification, summarization, and anomaly detection. They present many ways for describing and comparing time series, but they fo- cus on their values. This paper concentrates on a new as- pect - that of describing oscillation patterns. It presents a technique for time series similarity search, based on multi- ple temporal scales, defining a descriptor that uses the an- gular coefficients from a linear segmentation of the curve that represents the evolution of the analyzed series. Prelim- inary experiments with real datasets showed that our ap- proach correctly characterizes the oscillation of time series.