One of the most important tasks in audio signal processing is accurately identifying audio signals when they contain distortions from the real world. In this work, we suggest a scalable and lightweight framework for audio fingerprinting and matching that can recognize signals that have undergone different acoustic changes. The system uses a peak-based hashing mechanism to extract reliable time-frequency features from audio recordings. The spectral energy peaks and their temporal correlations are compactly represented by these fingerprints. Spectral statistics are used to create a low-dimensional embedding that further captures audio characteristics. We present a distortion augmentation pipeline that creates audio variations via dynamic range clipping, gain modification, additive noise, and equalization in order to evaluate robustness. Every version is fingerprinted separately and compared to a reference database made from the original audio files. Hash-set intersections and frequency-based score calculation are used for matching. The suggested fingerprinting technique is suitable for real-time media verification, speaker matching, and audio retrieval systems, as demonstrated by experimental results that show it maintains high identification accuracy under a variety of distortion conditions.