This study presents a novel vehicle environment detection technology that identifies driving conditions—including approaching vehicles, slippery roads, car horns, and emergency vehicle sirens—through ambient audio analysis. When the driver’s field of vision is limited or their attention is inattentive, it warns of sudden environmental changes to help the driver respond or increase the margin of reaction time. By extracting short-term energy (STE) and Mel-frequency cepstral coefficients (MFCCs) from environmental audio signals, the system captures both temporal and spectral features of surrounding sounds. These features are then fused using a channel attention mechanism, which enhances important feature channels and suppresses less relevant ones, improving recognition accuracy. The model leverages a residual structure to build a lightweight one-dimensional (1D) convolutional neural network (1D residual neural network [ResNet]), which effectively learns frequency shift features from moving sound sources, and can improve the recognition rate of dynamic targets and prevent degradation during deep learning, thereby being more flexible in responding to changes brought about by road conditions through dynamic feature weighting. This model provides a reliable auxiliary perception method when visual detection systems may fail, such as in extreme weather conditions. It helps both drivers and autonomous systems better identify environmental risks, thereby enhancing driving safety and system robustness.