Application of Time GAN-LSTM algorithm in constructing music aesthetic classification model

In the wave of the digital music era, music aesthetic classification has become the core issue of music information retrieval and personalized recommendation systems. The diversity and subjectivity of music aesthetics have brought great challenges to traditional classification methods, while the rise of deep learning technology has brought new opportunities to this field. This study proposes an algorithm that fuses temporal generative adversarial network (Time GAN) and long short-term memory network (LSTM), which is applied to construct a music aesthetic classification model in order to more accurately identify and classify music works. This study combines Time GAN and LSTM to realize a music aesthetic classification model. It provides a new perspective for the automatic classification of music aesthetics. In terms of experimental digital data, this study selected a music database containing 10,000 songs, which cover various styles such as classical, jazz, rock, and pop. In the experiment, we preprocessed these songs and extracted the Mel spectrogram of each song as input features. On this basis, we adopted Time GAN to generate additional training samples to enhance the generalization ability of the model. In the experiment, Time GAN successfully generated 5000 high-quality music samples, which, together with the raw data, constitute our training set. Through comparative experiments, we find that the Time GAN-LSTM model has achieved remarkable results in the task of music aesthetic classification. In the cross-validation test, the classification accuracy of the model reached 89.7%, which is a significant improvement compared to the 82.1% of LSTM alone and the 75.4% of traditional machine learning methods.