Audio Fingerprint Lab

MATLAB audio-processing pipeline for WAV ingestion, waveform analysis, FIR filtering, and spectrogram-based fingerprint matching.

Highlights

• Processes local WAV audio through a staged DSP pipeline.
• Extracts sample metadata, mono waveforms, FIR-filtered audio, filter responses, and compact audio fingerprints.
• Matches a query clip against a 9-song local database using hash-style spectrogram peak features.
• Keeps source audio, generated audio, and generated figures local-only so the public repository stays clean.

Features

Area	Status	Engineering Signal
WAV ingestion	Implemented	Uses MATLAB audio I/O with file validation and metadata extraction
Waveform analysis	Implemented	Converts stereo inputs to mono and exports time-domain plots
FIR filtering	Implemented	Applies 64-order low-pass, 32-order low-pass, and 64-order high-pass filters at 3 kHz
Response analysis	Implemented	Exports magnitude and phase response plots versus rad/sample and Hz
Fingerprint generation	Implemented	Builds spectrogram peak-pair hashes from local WAV clips
Fingerprint matching	Implemented	Scores query clips against a 9-song database with Jaccard similarity

Pipeline

WAV files
  -> ingest_audio_examples.m
  -> run_filter_pipeline.m
  -> run_fingerprint_matcher.m

The scripts are intentionally separate entry points so each processing stage can be run, validated, and extended independently.

Repository Structure

audio-fingerprint-lab/
|
├── data/
|   ├── examples/          local ingestion/filtering WAV inputs
|   └── recordings/        optional local recordings
|
├── src/
|   ├── ingest_audio_examples.m
|   ├── run_filter_pipeline.m
|   └── run_fingerprint_matcher.m
|
├── figures/               generated plots, ignored by Git
├── docs/
└── portfolio/

Additional local-only folders are created when running the full pipeline:

data/processed/
data/source_audio/
data/fingerprint_db/
data/queries/

Running The Pipeline

Use MATLAB from the repository root.

Ingestion and waveform export:

/Applications/MATLAB_R2025b.app/bin/matlab -batch "cd src; ingest_audio_examples"

FIR filtering and response plots:

/Applications/MATLAB_R2025b.app/bin/matlab -batch "assert(~isempty(which('fir1')), 'fir1 unavailable'); cd src; run_filter_pipeline"

Fingerprint matching:

/Applications/MATLAB_R2025b.app/bin/matlab -batch "assert(~isempty(which('spectrogram')), 'spectrogram unavailable'); cd src; run_fingerprint_matcher"

Local Audio Layout

The ingestion and filtering scripts expect:

data/examples/pirates.wav
data/examples/papas megalitriti.wav

The fingerprint matcher expects:

data/fingerprint_db/song01.wav
data/fingerprint_db/song02.wav
data/fingerprint_db/song03.wav
data/fingerprint_db/song04.wav
data/fingerprint_db/song05.wav
data/fingerprint_db/song06.wav
data/fingerprint_db/song07.wav
data/fingerprint_db/song08.wav
data/fingerprint_db/song09.wav
data/queries/query.wav

Local WAV files and generated artifacts are intentionally ignored by Git.

Implementation Notes

• FIR filters are designed with fir1 and Hamming windows.
• The 32-order low-pass filter demonstrates the wider transition band that comes from reducing filter order.
• Fingerprints are generated from local maxima in the magnitude spectrogram.
• Each fingerprint row stores [f1, f2, deltaT], linking two nearby spectral peaks.
• Query matching uses Jaccard similarity over unique fingerprint hashes.

Technical Stack

• MATLAB R2025b
• Signal Processing Toolbox
• WAV audio I/O
• FIR filter design
• Spectrogram analysis
• Local peak detection
• Hash-based similarity scoring

Local Data Policy

This repository is designed to be public without redistributing audio assets. Source audio, processed WAV files, generated figures, and private notes remain local-only.