Temporal Tensor Local Binary Pattern: A Novel Local Tensor Time Series Descriptor

Time Series is very common in both the industrial environment and real-life. Capturing the time dependency is very useful for time series analysis. Although the One-dimensional Local Binary Pattern can analyze the Univariate Time Series, it cannot effectively handle the Tensor Time Series with higher-order. Each variable in TTS depends not only on its past values but also has some dependency on other variables. In this paper, we propose a Temporal Tensor LBP operator to extract the discriminative temporal features from TTS by extending 1D-LBP operation from the scalar-wise to the tensor-wise. The TTLBP is simple yet discriminative, and can handle the tensor time series effectively. TTLBP is the first LBP variant for TTS analysis. We apply the proposed TTLBP in the edge intelligence-assisted Video Anomaly Detection system. Qualitative and quantitative comparisons demonstrate the effectiveness of the proposed TTLBP.