Robot interface classification based on CNNLSTM-Attention model

As the application of robots expands, the existing ground recognition method based on sound signals is easily affected by environmental noise, and vibration signals have become the research hotspot for complex ground environment recognition due to their stability. This paper explores the ground classification method for robots based on vibration signals, and proposes an innovative model combining convolutional neural network (CNN), long short-term memory network (LSTM), and attention mechanism to address the stability problem of robot movement on complex ground. Traditional ground recognition methods often use sound signals, but sound signals are easily affected by environmental noise, so ground classification based on vibration signals is more advantageous. CNN is used to extract local features of vibration signals, LSTM models the time-dependent relationship of signals, and the attention mechanism further improves the classification accuracy. Experimental results show that the improved CNN-LSTM model combined with the attention mechanism can effectively improve classification accuracy, especially in complex environments with different noise levels, showing strong stability and generalization ability.