Classification of Air Conditioner Sound Based on Mel Joint Features and Bidirectional Long Short-Term Memory Network

The air conditioner sound classification model can automatically identify abnormal sound of air conditioners, which is more efficient than the manual method in judging the quality of air conditioners. The key to sound classification is that the extracted features must have enough discrimination for different sound types. At the same time, the network used should conform to the characteristics of the data as much as possible. In this paper, a new classification model of air conditioner sound is proposed based on three types of sound data in SDU-Haier-ND data set. Firstly, the Mel spectral features and Mel-frequency cepstral coefficients of sound signals are extracted to form joint features. Then, the training feature set is input into bidirectional long short-term memory network for model training. Finally, the trained model is used to classify the test feature set. The experimental results show that the model has good classification ability.