mixboard-ios/Sources/Services/KeyDetector.swift

import Accelerate
import AVFoundation
import Foundation

/// Musical key detection using chromagram analysis and key profile matching.
struct KeyDetector {

    struct KeyResult {
        let key: String         // e.g. "C Major", "A Minor"
        let shortKey: String    // e.g. "C", "Am"
        let camelotCode: String // e.g. "8B", "1A"
        let confidence: Double  // 0.0 - 1.0
    }

    private static let noteNames = ["C", "C#", "D", "D#", "E", "F", "F#", "G", "G#", "A", "A#", "B"]

    // Krumhansl-Kessler major/minor profiles
    private static let majorProfile: [Double] = [6.35, 2.23, 3.48, 2.33, 4.38, 4.09, 2.52, 5.19, 2.39, 3.66, 2.29, 2.88]
    private static let minorProfile: [Double] = [6.33, 2.68, 3.52, 5.38, 2.60, 3.53, 2.54, 4.75, 3.98, 2.69, 3.34, 3.17]

    private static let camelotMajor = ["8B", "3B", "10B", "5B", "12B", "7B", "2B", "9B", "4B", "11B", "6B", "1B"]
    private static let camelotMinor = ["5A", "12A", "7A", "2A", "9A", "4A", "11A", "6A", "1A", "8A", "3A", "10A"]

    /// Detect the musical key of a track.
    static func detectKey(for track: Track) async throws -> KeyResult {
        let url = track.fileURL
        let file = try AVAudioFile(forReading: url)

        let format = file.processingFormat
        let sampleRate = format.sampleRate
        let maxFrames = AVAudioFrameCount(min(Double(file.length), sampleRate * 30))

        guard let buffer = AVAudioPCMBuffer(pcmFormat: format, frameCapacity: maxFrames) else {
            return KeyResult(key: "Unknown", shortKey: "?", camelotCode: "?", confidence: 0)
        }
        try file.read(into: buffer, frameCount: maxFrames)

        guard let floatData = buffer.floatChannelData else {
            return KeyResult(key: "Unknown", shortKey: "?", camelotCode: "?", confidence: 0)
        }

        let frameCount = Int(buffer.frameLength)
        var mono = [Float](repeating: 0, count: frameCount)
        let channels = Int(format.channelCount)
        for ch in 0..<channels {
            let ptr = floatData[ch]
            for i in 0..<frameCount { mono[i] += ptr[i] }
        }
        if channels > 1 {
            var div = Float(channels)
            vDSP_vsdiv(mono, 1, &div, &mono, 1, vDSP_Length(frameCount))
        }

        // Compute chromagram
        let fftSize = 8192
        let hopSize = 4096
        let log2n = vDSP_Length(log2(Double(fftSize)))
        guard let fftSetup = vDSP_create_fftsetup(log2n, FFTRadix(kFFTRadix2)) else {
            return KeyResult(key: "Unknown", shortKey: "?", camelotCode: "?", confidence: 0)
        }
        defer { vDSP_destroy_fftsetup(fftSetup) }

        let halfN = fftSize / 2
        var chromagram = [Double](repeating: 0, count: 12)
        var frameCount2 = 0

        var position = 0
        while position + fftSize <= frameCount {
            var frame = Array(mono[position..<position + fftSize])
            var window = [Float](repeating: 0, count: fftSize)
            vDSP_hann_window(&window, vDSP_Length(fftSize), Int32(vDSP_HANN_NORM))
            vDSP_vmul(frame, 1, window, 1, &frame, 1, vDSP_Length(fftSize))

            var real = [Float](repeating: 0, count: halfN)
            var imag = [Float](repeating: 0, count: halfN)
            frame.withUnsafeBufferPointer { fPtr in
                real.withUnsafeMutableBufferPointer { rPtr in
                    imag.withUnsafeMutableBufferPointer { iPtr in
                        var split = DSPSplitComplex(realp: rPtr.baseAddress!, imagp: iPtr.baseAddress!)
                        fPtr.baseAddress!.withMemoryRebound(to: DSPComplex.self, capacity: halfN) { cPtr in
                            vDSP_ctoz(cPtr, 2, &split, 1, vDSP_Length(halfN))
                        }
                        vDSP_fft_zrip(fftSetup, &split, 1, log2n, FFTDirection(kFFTDirection_Forward))
                    }
                }
            }

            var magnitudes = [Float](repeating: 0, count: halfN)
            real.withUnsafeBufferPointer { rPtr in
                imag.withUnsafeBufferPointer { iPtr in
                    var split = DSPSplitComplex(
                        realp: UnsafeMutablePointer(mutating: rPtr.baseAddress!),
                        imagp: UnsafeMutablePointer(mutating: iPtr.baseAddress!))
                    vDSP_zvabs(&split, 1, &magnitudes, 1, vDSP_Length(halfN))
                }
            }

            // Map frequency bins to pitch classes
            for bin in 1..<halfN {
                let freq = Double(bin) * sampleRate / Double(fftSize)
                guard freq > 60 && freq < 5000 else { continue }
                let midiNote = 69.0 + 12.0 * log2(freq / 440.0)
                let pitchClass = Int(round(midiNote)) % 12
                let normalizedClass = (pitchClass + 12) % 12
                chromagram[normalizedClass] += Double(magnitudes[bin])
            }
            frameCount2 += 1
            position += hopSize
        }

        guard frameCount2 > 0 else {
            return KeyResult(key: "Unknown", shortKey: "?", camelotCode: "?", confidence: 0)
        }

        for i in 0..<12 { chromagram[i] /= Double(frameCount2) }

        // Match against key profiles using Pearson correlation
        var bestCorrelation = -Double.greatestFiniteMagnitude
        var bestKey = 0
        var bestIsMajor = true

        for root in 0..<12 {
            let rotated = (0..<12).map { chromagram[($0 + root) % 12] }
            let majorCorr = pearsonCorrelation(rotated, majorProfile)
            let minorCorr = pearsonCorrelation(rotated, minorProfile)

            if majorCorr > bestCorrelation {
                bestCorrelation = majorCorr
                bestKey = root
                bestIsMajor = true
            }
            if minorCorr > bestCorrelation {
                bestCorrelation = minorCorr
                bestKey = root
                bestIsMajor = false
            }
        }

        let confidence = max(0, min(1, (bestCorrelation + 1) / 2))
        let noteName = noteNames[bestKey]
        let keyName = bestIsMajor ? "\(noteName) Major" : "\(noteName) Minor"
        let shortKey = bestIsMajor ? noteName : "\(noteName)m"
        let camelot = bestIsMajor ? camelotMajor[bestKey] : camelotMinor[bestKey]

        return KeyResult(key: keyName, shortKey: shortKey, camelotCode: camelot, confidence: confidence)
    }

    private static func pearsonCorrelation(_ a: [Double], _ b: [Double]) -> Double {
        let n = Double(a.count)
        let sumA = a.reduce(0, +)
        let sumB = b.reduce(0, +)
        let sumAB = zip(a, b).map(*).reduce(0, +)
        let sumA2 = a.map { $0 * $0 }.reduce(0, +)
        let sumB2 = b.map { $0 * $0 }.reduce(0, +)

        let numerator = n * sumAB - sumA * sumB
        let denominator = sqrt((n * sumA2 - sumA * sumA) * (n * sumB2 - sumB * sumB))
        guard denominator > 0 else { return 0 }
        return numerator / denominator
    }
}