Odor identification optimization via DSCNN-MSA

Odor identification is essential in applications such as environmental monitoring, food safety, and medical diagnostics. This study proposes an optimized deep learning model, DSCNN_MSA, which combines Depthwise Separable Convolutional Neural Networks (DSCNN) with Multi-Head Self-Attention (MSA) to enhance feature extraction and classification performance. The DSCNN improves computational efficiency, while MSA captures long-range dependencies for better odor representation. Experimental results on an electronic nose dataset show that DSCNN-MSA achieves an average accuracy of 94.29% over ten runs, outperforming conventional models. This approach provides a more effective and efficient solution for machine olfaction systems.