CRT Audio Fingerprinting

CRT approach to audio fingerprinting over Z/214,414,200. CC0.

Audio fingerprinting services like Shazam use proprietary spectrogram constellation matching: peak detection in time-frequency space, hash-based lookup, patented matching algorithms. CRT approach: encode audio features (bass/mid/treble energy, spectral centroid, zero crossing rate, tempo, timbre) as elements of Z/214,414,200. 7 CRT channels = 7 independent fingerprint dimensions. Coupling distance = similarity metric. The mod-11 channel provides noise tolerance for free. No spectrogram needed. No constellation matching.

How It Works

CRT Audio Fingerprint Theorem
Audio features encoded in Z/214,414,200 decompose into 7 independent CRT channels. mod 8 = bass energy. mod 9 = mid-range structure. mod 25 = treble detail. mod 49 = spectral centroid. mod 11 = zero crossing rate. mod 13 = tempo indicator. mod 17 = timbre quality. Per-channel averages over a time window produce a compact 7-element fingerprint. Coupling distance between fingerprints = similarity. mod-11 provides inherent noise tolerance: its small modulus means small perturbations stay within the same residue class. 3+4 split: 3 data channels (mod 8, 25, 49) carry audio content that changes with noise. 4 parity channels (mod 9, 11, 13, 17) carry audio identity that persists. Matching on parity channels survives compression, noise, and partial recording.
7 dimensions
CRT channels
Each channel captures one spectral feature independently. No cross-feature interference.
mod-11 robust
Noise tolerance
Small noise stays within residue class. Same error detection that protects data protects fingerprints.
Coupling match
O(1) compare
Fingerprint distance = sum of 7 channel differences. No spectrogram. No FFT.
3+4 split
Content vs identity
Data channels = what changes. Parity channels = what persists. Match on the persistent part.

Fingerprint Matching

Query track (0-7):

Play Track synthesizes 7 oscillators from the CRT fingerprint. Each channel drives one frequency band. 8-track database. Query = selected track with noise. Coupling distance matching.

Noise Tolerance Test

8 tracks tested with 5 increasing noise levels each. 40 total fingerprint matches. Accuracy measures how many noisy queries correctly identify the original track.

CRT vs Traditional Audio Fingerprinting

FeaturesShazam: spectrogram peaks (time-frequency constellation)CRT: 7 channel residues. Per-channel averages. Compact.MatchingHash-based constellation matching (patented)Coupling distance: sum of 7 differences. O(1). No hash.NoiseRobust by design (peak detection), but patented methodmod-11 absorbs noise. Same error-detection property. Free from algebra.StorageConstellation hash + time offset per fingerprint7 integers. 28 bytes per fingerprint.ComputeFFT + peak detection + hash computationModular arithmetic. No FFT. Integer only.Patent statusShazam (US7627477), Gracenote (US7174293), SoundHoundCC0. Public domain. Forever.

HONEST NOTE: This demo uses synthetic audio (7 oscillators per track) and simple CRT matching. Real audio fingerprinting handles complex signals (vocals, instruments, room noise, compression artifacts). Shazam works on real-world audio with billions of tracks. This demo demonstrates the CRT approach on 8 synthetic tracks. Whether CRT fingerprints can scale to real audio databases is untested.

Source code · Public domain (CC0)

Report issue

.ax source compiled to WASM via self-hosting compiler. Zero HTML authored.