@playground-sessions/basic-pitch

A sandbox implementation of three different pitch detection algorithms for comparison:

1. Autocorrelation (time-domain)
2. Average Magnitude Difference Function (AMDF)
3. Fast Fourier Transform (FFT) with Harmonic Product Spectrum

This is kind of a failure. I didn't get it to work.

Features

• Three different pitch detection methods in one package
• Compare algorithms side-by-side
• Interactive demo with real-time visualization
• TypeScript types included
• Zero dependencies

Installation

npm install @playground-sessions/basic-pitch

Usage

import {
  AutocorrelationPitchDetector,
  AmdfPitchDetector,
  FftPitchDetector,
  CombinedPitchDetector
} from '@playground-sessions/basic-pitch';

// Use individual detectors
const detector = new AutocorrelationPitchDetector({
  frameSize: 2048,
  sampleRate: 44100,
  minFrequency: 50,
  maxFrequency: 2000
});

// Or use the combined detector that picks the best result
const combined = new CombinedPitchDetector({
  frameSize: 2048,
  sampleRate: 44100
});

// Analyze audio frames
const frame = new Float32Array(2048); // Your audio data here
const result = detector.analyze(frame);
console.log(`Frequency: ${result.frequency} Hz`);
console.log(`Confidence: ${result.confidence}`);
console.log(`Method: ${result.method}`);

Demo

An interactive demo is included that lets you compare all three methods in real-time:

1. Clone this repository
2. Run npm install
3. Run npm run dev
4. Open your browser and try the pitch detector

How it Works

Autocorrelation Method

• Compares the signal with delayed versions of itself
• Finds periodic patterns in the time domain
• Good for clean, monophonic signals

AMDF Method

• Similar to autocorrelation but uses differences instead of products
• Can be more robust to noise in some cases
• Computationally efficient

FFT Method

• Transforms the signal to the frequency domain
• Uses Harmonic Product Spectrum for better accuracy
• Good for complex tones with strong harmonics

Options

All detectors accept these options:

• frameSize: Size of the analysis frame (default: 2048)
• sampleRate: Sample rate of the audio (default: 44100)
• minFrequency: Minimum frequency to detect (default: 50 Hz)
• maxFrequency: Maximum frequency to detect (default: 2000 Hz)

Contributing

This is an educational project built during the WebAudio Pitch Jam '25. Contributions are welcome!