| 1 | import Foundation |
| 2 | |
| 3 | /// One committed "intentional" note — a stable pitch you actually meant to sing, |
| 4 | /// with a start, an end, and how sharp/flat you held it. |
| 5 | struct NoteEvent: Identifiable { |
| 6 | let id: Int |
| 7 | let midi: Int // integer MIDI (concert pitch, before transpose) |
| 8 | let onset: Double // seconds on the CACurrentMediaTime clock |
| 9 | var offset: Double // seconds |
| 10 | let meanCents: Double // signed average deviation from the integer, in cents |
| 11 | var duration: Double { offset - onset } |
| 12 | } |
| 13 | |
| 14 | /// Turns a stream of fractional-MIDI pitch samples into discrete note events — |
| 15 | /// the piano-roll layer that sits above the raw pitch contour. |
| 16 | /// |
| 17 | /// This is a lightweight, real-time analogue of pYIN's note HMM: |
| 18 | /// • a tolerance band around the current note absorbs vibrato and small drift, |
| 19 | /// so a wobbling held note reads as ONE note instead of flickering; |
| 20 | /// • when the pitch leaves that band the note is closed and a new one opens; |
| 21 | /// • notes shorter than `minDuration` are discarded, so scoops and passing |
| 22 | /// tones on the way to a target don't become spurious notes; |
| 23 | /// • a silence gap longer than `gapToEnd` closes the held note. |
| 24 | final class NoteSegmenter { |
| 25 | // Tunables. Cents are hundredths of a semitone; a semitone is 100 cents, so |
| 26 | // a note "owns" ±centsTolerance around its center — the overlap past the |
| 27 | // 50-cent midpoint is the hysteresis that keeps vibrato from splitting notes. |
| 28 | private let centsTolerance = 62.0 |
| 29 | private let minDuration = 0.10 // seconds; below this a region is dropped |
| 30 | private let gapToEnd = 0.13 // seconds of silence that ends a held note |
| 31 | private let retain = 30.0 // seconds of committed notes kept |
| 32 | |
| 33 | private(set) var committed: [NoteEvent] = [] |
| 34 | |
| 35 | // In-progress note. |
| 36 | private var pendingMidi: Int? |
| 37 | private var pendingOnset = 0.0 |
| 38 | private var pendingCentsSum = 0.0 |
| 39 | private var pendingCount = 0 |
| 40 | private var lastVoiced = 0.0 |
| 41 | private var nextID = 0 |
| 42 | |
| 43 | /// Feed one voiced sample. Returns true iff a note was committed (ended). |
| 44 | @discardableResult |
| 45 | func feed(time: Double, midi: Double) -> Bool { |
| 46 | var didCommit = false |
| 47 | |
| 48 | // A silence gap since the previous voiced sample closes the held note. |
| 49 | if pendingMidi != nil, time - lastVoiced > gapToEnd { |
| 50 | didCommit = closePending(offset: lastVoiced) |
| 51 | } |
| 52 | |
| 53 | if let cur = pendingMidi { |
| 54 | let distCents = abs(midi - Double(cur)) * 100 |
| 55 | if distCents <= centsTolerance { |
| 56 | // Still the same note — absorb the sample (this is what swallows |
| 57 | // vibrato even as the nearest semitone flips back and forth). |
| 58 | pendingCentsSum += (midi - Double(cur)) * 100 |
| 59 | pendingCount += 1 |
| 60 | lastVoiced = time |
| 61 | } else { |
| 62 | // Pitch has moved off the note: close it, open a new one. |
| 63 | didCommit = closePending(offset: lastVoiced) || didCommit |
| 64 | openPending(time: time, midi: midi) |
| 65 | } |
| 66 | } else { |
| 67 | openPending(time: time, midi: midi) |
| 68 | } |
| 69 | return didCommit |
| 70 | } |
| 71 | |
| 72 | /// Feed a detector "silence" tick so a held note eventually closes even if no |
| 73 | /// further voiced sample arrives. (Gaps are also caught lazily in `feed`.) |
| 74 | @discardableResult |
| 75 | func feedSilence(now: Double) -> Bool { |
| 76 | guard pendingMidi != nil, now - lastVoiced > gapToEnd else { return false } |
| 77 | return closePending(offset: lastVoiced) |
| 78 | } |
| 79 | |
| 80 | /// The note currently being sung, as a provisional event for live rendering. |
| 81 | var pending: NoteEvent? { |
| 82 | guard let m = pendingMidi, pendingCount > 0 else { return nil } |
| 83 | // Only surface it once it's plausibly a note, not a passing scoop. |
| 84 | guard lastVoiced - pendingOnset >= minDuration * 0.5 else { return nil } |
| 85 | return NoteEvent(id: -1, midi: m, onset: pendingOnset, offset: lastVoiced, |
| 86 | meanCents: pendingCentsSum / Double(pendingCount)) |
| 87 | } |
| 88 | |
| 89 | func reset() { |
| 90 | committed.removeAll(keepingCapacity: true) |
| 91 | pendingMidi = nil |
| 92 | pendingCount = 0 |
| 93 | } |
| 94 | |
| 95 | // MARK: - Private |
| 96 | |
| 97 | private func openPending(time: Double, midi: Double) { |
| 98 | let m = Int(midi.rounded()) |
| 99 | pendingMidi = m |
| 100 | pendingOnset = time |
| 101 | pendingCentsSum = (midi - Double(m)) * 100 |
| 102 | pendingCount = 1 |
| 103 | lastVoiced = time |
| 104 | } |
| 105 | |
| 106 | @discardableResult |
| 107 | private func closePending(offset: Double) -> Bool { |
| 108 | guard let m = pendingMidi else { return false } |
| 109 | pendingMidi = nil |
| 110 | guard offset - pendingOnset >= minDuration, pendingCount > 0 else { return false } |
| 111 | committed.append(NoteEvent(id: nextID, midi: m, onset: pendingOnset, |
| 112 | offset: offset, meanCents: pendingCentsSum / Double(pendingCount))) |
| 113 | nextID += 1 |
| 114 | let cutoff = offset - retain |
| 115 | if let first = committed.first, first.offset < cutoff { |
| 116 | committed.removeAll { $0.offset < cutoff } |
| 117 | } |
| 118 | return true |
| 119 | } |
| 120 | } |