Acoustic scene classification method, model training method, deployment method, model, chip, device, electronic equipment and storage medium

The invention relates to the technical field of artificial intelligence, and discloses an acoustic scene classification method, a model training method, a deployment method, a model, a chip, an electronic device and a storage medium, the acoustic scene classification model comprises a data acquisition module, an audio signal preprocessing module, a plurality of depth expansion residual blocks connected in sequence, a convolution layer and a full connection layer, by adopting the deep expansion residual block, the network receptive field can be improved on the premise of not increasing network parameters and calculation amount, the global information capturing capability of the network is higher, gradient disappearance possibly occurring in the network training process can be avoided by combining residual operation with expansion convolution, and the network training efficiency is improved through instance normalization processing. According to the technical scheme, the frequency domain information and the tim