PhonemeAnalyzer

PhonemeAnalyzer.ts

import { PhonemeKey } from '../types';

/**
* PhonemeAnalyzer redesigned for the Hanna-Barbera animation system (Preston Blair).
* Classifies audio input into 6 mouth shapes (A-F) based on formants and amplitude.
*/
export class PhonemeAnalyzer {
  private analyser: AnalyserNode;
  private dataArray: Uint8Array;
  private binFrequency: number;
  
  private lastPhoneme: PhonemeKey = 'A';
  private phonemeHoldFrames: number = 0;
  private readonly MIN_HOLD_FRAMES = 1; 
  
  constructor(audioContext: AudioContext, source: AudioNode) {
    this.analyser = audioContext.createAnalyser();
    this.analyser.fftSize = 512; 
    this.analyser.smoothingTimeConstant = 0.35;
    source.connect(this.analyser);
    
    this.dataArray = new Uint8Array(this.analyser.frequencyBinCount);
    this.binFrequency = audioContext.sampleRate / this.analyser.fftSize;
  }

  private hzToBin(hz: number): number {
    return Math.round(hz / this.binFrequency);
  }

  private getAverageFrequencyRangeHz(startHz: number, endHz: number): number {
    const startBin = this.hzToBin(startHz);
    const endBin = this.hzToBin(endHz);
    let sum = 0;
    const start = Math.max(0, startBin);
    const end = Math.min(this.dataArray.length, endBin);
    const count = end - start;
    if (count <= 0) return 0;
    for (let i = start; i < end; i++) sum += this.dataArray[i];
    return sum / count / 255;
  }

  public getPhoneme(): PhonemeKey {
    const detected = this.detectHannaBarberaShape();
    
    if (detected === this.lastPhoneme) {
      this.phonemeHoldFrames = 0;
      return detected;
    }
    
    this.phonemeHoldFrames++;
    
    if (detected === 'D' || detected === 'A' || this.phonemeHoldFrames >= this.MIN_HOLD_FRAMES) {
      this.lastPhoneme = detected;
      this.phonemeHoldFrames = 0;
    }
    
    return this.lastPhoneme;
  }

  private detectHannaBarberaShape(): PhonemeKey {
    this.analyser.getByteFrequencyData(this.dataArray);
    
    const energyLow = this.getAverageFrequencyRangeHz(80, 400);    
    const energyMid = this.getAverageFrequencyRangeHz(400, 1500);  
    const energyHigh = this.getAverageFrequencyRangeHz(2000, 6000);
    
    const totalVolume = (energyLow + energyMid + energyHigh) / 3;
    
    // If there is silence, we return 'A' (Rests/Closed)
    if (totalVolume < 0.05) return "A";

    if (totalVolume < 0.12 && energyLow > energyHigh) return "A";
    if (totalVolume > 0.4 && energyMid > energyHigh * 1.5) return "D";
    if (energyHigh > energyMid * 0.8 || (energyHigh > 0.2 && energyLow < 0.3)) return "B";
    if (energyLow > energyMid * 2.5 && energyHigh < 0.1) return "F";
    if (energyLow > energyHigh * 2 && energyMid > energyHigh) return "E";

    return "C";
  }

  public dispose() {
    try { this.analyser.disconnect(); } catch (e) {}
  }
}