Audio classification method, system and device based on feature fusion

The invention provides an audio classification method, system and device based on feature fusion, and relates to the technical field of audio scene classification, and the method mainly comprises the steps: obtaining audio scene data, and carrying out the preprocessing, and obtaining an audio signal; the audio signal is converted into a Mel spectrogram, the Mel spectrogram is divided into two parts according to the proportion, one part serves as a training set, and the other part serves as a test set; inputting the training set into a fusion model for training, optimizing model parameters by fusing data features, and adjusting hyper-parameters to obtain data tags; inputting the test set into the fusion model for calculation to obtain a classification label; based on the classification labels, evaluation calculation is carried out, and the classification accuracy is obtained. According to the scheme, the features are identified through a plurality of deep learning models and fused to obtain the optimized fusio