The paper discusses a proposed AI and ML integrated audio fingerprinting algorithm to improve accuracy. The research builds on the Dejavu Project’s foundations and emphasizes real-world scenario simulations with diverse background noises and distortions. The study highlights the performance of the new system, which shows 100% accuracy within a 5-second audio input, and discusses the practical implications of storage space and speed trade-offs. The aim is to enhance the adaptability of audio fingerprinting to address challenges in different environments and applications.
Publication date: 23 Feb 2024
Project Page: https://github.com/worldveil/dejavu
Paper: https://arxiv.org/pdf/2402.13957
