flâneur

How does Audio Fingerprinting work - Emysound

emysound.com · 5,318 words · saved by 1 readers

I have been developing the SoundFingerprinting open source project for the last ten years. One of the questions I often receive is “how does music recognition works?” For the library users, it is somewhat similar to a one-way hash function. You provide a file at the input, and after a certain number of conversions, you get “audio fingerprints” at the output. Looking at the actual values of these fingerprints, they are entirely opaque. The question begs: What information do these integers contain? My goal in this article is to dissect the algorithm’s two main steps: compression and hashing. Additionally, I will explain why audio and image search are so intertwined. Compression - from Latin comprimere press together From a higher level, all audio fingerprinting algorithms go through two transformation steps: lossy compression and hashing. We want to preserve as much relevant information as possible with the smallest footprint. Look at the following array: We need at least 40 bytes to sto

How does Audio Fingerprinting work - Emysound Audio Fingerprinting I have been developing the SoundFingerprinting open source project for the last ten years. One of the questions I often receive is “how does music recognition works?” For the library users, it is somewhat similar to a one-way hash function. You provide a file at the input, and after a certain number of conversions, you get “audio fingerprints” at the output. Looking at the actual values of these fingerprints, they are entirely opaque. 287121152, 1275791411, 539499396, 185209916, 319820588, 1244475492, 1751526233, 1862426472, 30

saved by

related reading