We propose in this paper a classification framework based on long short-term memory (LSTM) networks. To achieve robust classification, a one dimension-aggregate approximation (1d-AX) is employed to ...