lydsøk

Lydsøk, or sound search, refers to techniques and technologies for locating audio content in a database or on the web by using an audio query rather than text. It can operate with short audio clips, hummed melodies, or descriptive metadata, and is used in media libraries, streaming platforms, and broadcast monitoring. In Norwegian contexts, lydsøk commonly denotes these methods and their applications.

Core techniques include acoustic fingerprinting, which encodes stable features of an audio signal into compact codes for fast matching, and feature-based search that analyzes spectral, harmonic, and rhythmic properties. Query-by-example allows users to submit an audio sample or a hummed tune. Many modern systems combine these approaches with machine learning to improve robustness to noise, compression, and distortion, and to support cross-modal or cross-language queries.

Applications span music identification and rights management, podcast and broadcast search, audio-based discovery in digital libraries, and surveillance or safety monitoring. For users, lydsøk enables finding a track or clip without exact metadata; for rights holders, it aids monitoring and licensing.

Limitations include sensitivity to noise and overlapping sounds, distortion from encoding, and the need for large, well-curated reference databases. Privacy and copyright considerations also shape deployment, particularly in public or monitored environments.

Historically, practical audio fingerprinting and music identification emerged in the late 1990s and early 2000s, with widespread adoption in the following decades. See also content-based audio retrieval, audio fingerprinting, and query by humming.