A Parallelized Data Stream Processing System Using Dynamic Time Warping Distance

Due to the development of sensing technologies, sensor stream processing systems have been getting great attention. Sensor stream processing systems enable us to recognize the state of objects. For example, attach an acceleration sensor to a man and the system beforehand stores sample sensor data for the manpsilas walking and standing states. By comparing the current data stream with the stored sample data, the system can recognize whether he is walking or standing. Recognizable states increases in proportion to the number of the sample data. However, the processing time lengthens as the number of sample data increases. Long processing time is impractical for data stream processing systems. In this paper, we design and implement a data stream processing system using dynamic time warping (DTW) distance. DTW is robust for changes in time and often used for streaming data analysis. In our implemented system, we reduce the processing time by parallelizing the calculation of the DTW distance.