mirror of
https://github.com/9x/sheetmusictrainer.git
synced 2026-11-03 22:03:54 +01:00
Jens Mohrmann
079d574b66
- Pick detector per instrument range: ACF2PLUS (robust down to ~40 Hz) for normal instruments, YIN only for the whistle register. pitchfinder's YIN latches onto tau=2 (~20 kHz garbage) below ~100 Hz and for quiet input; the old fixed 4096 window masked this, the smaller window exposed it. - RMS-normalize analysis input for the YIN path only; ACF2PLUS keeps the raw signal so its internal unpitched gate rejects quiet decayed audio instead of amplifying noise into confident wrong pitches (tuner stuck on 'a note'). - Adaptive analysis window sized to the instrument's lowest note (~2 periods): half the latency, a quarter of the CPU for guitar/piano/voice; bass keeps the full window for E1. - Reject detections implausibly below the instrument floor (mains hum). - Clear stale pitch readings after 200 ms instead of keeping them forever. - Shared AudioEngine singleton for all playback (was: two AudioContexts); tracks audibility so the app stops hearing its own speaker (ear-training self-scoring) and gates matching plus display during playback. - MatchTracker: grace-period based note hold instead of reset-on-every-flicker (YIN flickers during attacks/decays; the old hard reset made the 50 ms threshold fire unreliably). - Display damping in the producer (EMA for cents/level, 3-frame note-name hysteresis); display components render directly without CSS transitions. - Regression tests: synthetic signals through the exact pipeline helpers (MatchTracker + pitchDetection test suites).
89 lines
4.3 KiB
TypeScript
89 lines
4.3 KiB
TypeScript
/**
|
||
* Global application configuration
|
||
*/
|
||
|
||
// How long (in ms) the target note must be detected to count as correct.
|
||
// Protects against spurious single-frame detections.
|
||
export const NOTE_MATCH_THRESHOLD_MS = 50;
|
||
|
||
// Grace period (in ms): brief detection dropouts or wrong frames while holding
|
||
// a note do NOT reset the match timer. Pitch detectors (YIN) routinely flicker
|
||
// for a frame or two during note attacks and decays; a hard reset on every
|
||
// flicker makes the threshold above feel broken ("shows green but won't register").
|
||
export const NOTE_MATCH_GRACE_MS = 150;
|
||
|
||
// How long (in ms) after speaker playback ends that mic-based matching stays
|
||
// gated, so the app doesn't "hear itself" (ear-training auto-play, Play Note
|
||
// button, virtual instruments).
|
||
export const PLAYBACK_GATE_TAIL_MS = 400;
|
||
|
||
// Safety cap for the "wait for silence" option: if the mic never goes quiet
|
||
// (continuously detectable sound in the room), the next reference note plays
|
||
// anyway after this delay so the app can't stall indefinitely.
|
||
export const REFERENCE_QUIET_MAX_WAIT_MS = 8000;
|
||
|
||
// Display smoothing (exponential moving average factor, 0-1) applied to the
|
||
// cents reading and the audio level in the detection loop (~60 fps). Keeps
|
||
// the tuner needle and level meter readable without CSS transitions — a
|
||
// transition on a value that changes every frame is interrupted 60x per
|
||
// second, which Safari/WebKit handles by leaving the element stuck.
|
||
// 0.25 ≈ 80ms effective smoothing: damps jitter but stays very responsive.
|
||
export const DISPLAY_EMA_FACTOR = 0.25;
|
||
|
||
// How many consecutive detection frames a different note must persist before
|
||
// the displayed note name switches (~50ms at 60fps). Prevents the note name
|
||
// from flickering between adjacent notes when playing right at a semitone
|
||
// boundary. Matching is unaffected (it uses the raw midi value).
|
||
export const NOTE_DISPLAY_STABLE_FRAMES = 3;
|
||
|
||
// How long (in ms) the last detected pitch is kept on screen after the signal
|
||
// disappears, before the readout is cleared. Prevents flicker between frames
|
||
// without leaving stale notes hanging forever.
|
||
export const PITCH_READING_HOLD_MS = 200;
|
||
|
||
// Pitch analysis window (samples @ AudioContext rate).
|
||
// The analyzer adapts the window down from BUFFER_SIZE based on the lowest
|
||
// frequency that must be detected (~2 periods minimum). Smaller window =
|
||
// lower latency and 4x less CPU per analysis. 4096 covers bass E1 (~41 Hz).
|
||
export const PITCH_BUFFER_SIZE = 4096;
|
||
export const PITCH_MIN_WINDOW = 1024;
|
||
|
||
// Fallback for the lowest detectable frequency before the app configures
|
||
// the actual instrument range. ~70 Hz covers guitar low E with margin.
|
||
export const PITCH_DEFAULT_LOWEST_HZ = 70;
|
||
|
||
// Instruments whose lowest note is at/above this frequency use YIN instead of
|
||
// ACF2PLUS: ACF2PLUS picks subharmonics above ~2 kHz, while YIN's weakness
|
||
// below ~100 Hz is irrelevant for high registers (e.g. whistle).
|
||
export const PITCH_HIGH_REGISTER_HZ = 300;
|
||
|
||
// Target RMS for normalizing the analysis buffer before pitch detection
|
||
// (YIN/high-register path only — see selectDetector in PitchAnalyzer).
|
||
// pitchfinder's YIN force-sets yinBuffer[1] = 1 before normalization, so at
|
||
// low absolute amplitudes its difference function gets dwarfed and it latches
|
||
// onto tau=2 (~20 kHz garbage). A constant operating level fixes this.
|
||
export const PITCH_NORMALIZE_TARGET_RMS = 0.3;
|
||
|
||
// Detected pitches below (instrument floor × this factor) are rejected as
|
||
// implausible — e.g. 50/60 Hz mains hum showing up as a confident wrong note
|
||
// while a guitar (floor ~78 Hz) is not playing. ~4 semitones of margin.
|
||
export const PITCH_BELOW_FLOOR_REJECT_FACTOR = 2 ** (-4 / 12); // ≈ 0.794
|
||
|
||
// Microphone Sensitivity Configuration
|
||
// Sensitivity Scale: 0.0 (Low) to 1.0 (High)
|
||
// Linear mapping to dB thresholds
|
||
export const MIC_SENSITIVITY_DB_RANGE = {
|
||
min: -10, // 0.0 Sensitivity: Loudest input required (e.g. loud singing/instrument)
|
||
max: -100 // 1.0 Sensitivity: Quietest input accepted (mic floor)
|
||
};
|
||
|
||
export const MIC_DEFAULT_SENSITIVITY = 0.75; // Bias towards high sensitivity for ease of use
|
||
|
||
// Audio Constraints for getUserMedia
|
||
// Disabled processing is generally better for musical pitch detection
|
||
export const AUDIO_CONSTRAINTS = {
|
||
echoCancellation: false,
|
||
noiseSuppression: false,
|
||
autoGainControl: false,
|
||
};
|