Environmental Sound Classification Using 1-D and 2-D Convolutional Neural Networks

Audio classification also called as sound classification is used to recognize and distinguish between the different types of environmental sounds. Audio classification tasks can involve classifying audio recordings into different categories, such as genres, emotions, or states. This can be used to identify a particular song or speech, classify audio recordings for specific applications, and gain a better understanding of the structure of audio data. Deep learning has recently become one of the most effective methods for categorizing audio, and numerous promising methods have been put forth. Existing systems have used three distinct types of time-frequency representation and various algorithms to classify sounds. In the model presented, features are initially extracted from audio signals through Mel-Frequency Cepstral Coefficients methods applied in both time and frequency domains. The proposed environmental sound classification system utilizes deep-learning techniques such as Convolutional Neural Networks with 1D and 2D convolutions, yielding accuracy values of 93.02% and 92.87% respectively.