Here’s the actual pipeline: 1. Audio capture → spectrogram The incoming audio signal is sampled and transformed using a Short-Time Fourier Transform (STFT). This converts the waveform (time domain) into a spectrogram (time vs frequency vs intensity). Instead of raw sound, Shazam now works with frequency patterns over time. 2. Peak detection The system identifies high-energy frequency points (local maxima) in the spectrogram. These peaks are robust to noise, compression, and background sounds, which is why Shazam works in loud environments. 3. Fingerprint generation Rather than storing the full spectrogram, Shazam creates “fingerprints” by pairing peaks. Each pair encodes: • Frequency of point A • Frequency of point B • Time difference between them This produces a sparse but highly distinctive representation of the audio. 4. Hashing and indexing Each fingerprint is converted into a hash (a compact numerical key). These hashes are precomputed for millions of songs and stored in ...