“…Therefore, an algorithm capable of effectively classifying sequential data of varying lengths is necessary for SSR. Various algorithms have been employed for sequence classification tasks, including dynamic time warping (DTW) [44], [45], [46], [47], [48], long short-term memory (LSTM) [49], [50], bidirectional long short-term memory (BLSTM) [51], [52], Gaussian mixture model-hidden Markov model (GMM-HMM) [53], [54], [55], [56], [57], and deep neural network-hidden Markov model (DNN-HMM) [58], [59].…”