Effective communication is critical in rescue operations, particularly in hazardous situations such as caverns, for the safety and coordination of victims and rescuers. This study investigates the application of deep learning techniques to improve audio communication using a Wiener filter for noise reduction. The Wiener filter significantly reduces non-human noise, enhancing the clarity of human speech. The system can effectively identify and classify human speech even in noisy situations using deep learning-based audio classification. The Wiener filter significantly improves the signal-to-noise ratio (SNR) and mean squared error (MSE). This approach highlights the ability of advanced audio processing techniques to improve communication efficiency and efficacy in rescue operations.