Time Series Segmentation Using Neural Networks with Cross-Domain Transfer Learning

Searching for characteristic patterns in time series is a topic addressed for decades by the research community. Conventional subsequence matching techniques usually rely on the definition of a target template pattern and a searching method for detecting similar patterns. However, the intrinsic variability of time series introduces changes in patterns, either morphologically and temporally, making such techniques not as accurate as desired. Intending to improve segmentation performances, in this paper, we proposed a Mask-based Neural Network (NN) which is capable of extracting desired patterns of interest from long time series, without using any predefined template. The proposed NN has been validated, alongside a subsequence matching algorithm, in two datasets: clinical (electrocardiogram) and human activity (inertial sensors). Moreover, the reduced dimension of the data in the latter dataset led to the application of transfer learning and data augmentation techniques to reach model convergence. The results have shown the proposed model achieved better segmentation performances than the baseline one, in both domains, reaching average Precision and Recall scores of 99.0% and 97.5% (clinical domain), along with 77.0% and 71.4% (human activity domain), introducing Neural Networks and Transfer Learning as promising alternatives for pattern searching in time series.

A Real Time Localization System for Vehicles Using Terrain-Based Time Series Subsequence Matching

Global Navigation Satellite Systems (GNSSs) are commonly used for positioning vehicles in open areas. Yet a GNSS frequently encounters loss of lock in urban areas. This paper presents a new real-time localization system using measurements from vehicle odometer data and data from an onboard inertial measurement unit (IMU), in the case of lacking GNSS information. A Dead Reckoning model integrates odometer data, IMU angular and velocity data to estimate the rough position of the vehicle. We then use an R-Tree structured reference road map of pitch data to boost spatial search efficiency. An optimized time series subsequence matching method matches the measured pitch data and the stored pitch data in reference road map for more accurate position estimation. The two estimated positions are fused using an extended Kalman filter model for final localization. The proposed localization system was tested for computational complexity with a median runtime of 12 ms, and for positioning accuracy with a median position error of 0.3 m.