Algorithm 1: LSTM Based Sensory Landmark Recognition.

Input: labeled train data set $\left\{X^{Train},L^{Train}\right\}$ and unlabeled test data set $\left\{X^{Test}\right\}$ Output: predicted sensory landmark labels of the test data set $L^{Pre}$ 1 Collect data 2 Process data by interpolation, filtering, segmentation, and labeling 3 Initialize neural network parameters randomly 4 Train and compute the parameters 5 Use the trained neural network to predict the sensory landmark labels of the data set $L^{Pre}$