Add FLAC audio quality analysis and spectrum visualization (#110)

Introduces backend support for analyzing FLAC audio files, including technical metrics and frequency spectrum extraction. Adds frontend components and hooks for file selection, analysis, and visualization, integrating a new Audio Quality Analyzer dialog into the UI. Updates types and dependencies to support audio analysis features.
2025-11-25 22:05:12 +01:00
parent ddf1844237
commit 2172981110
12 changed files with 1145 additions and 0 deletions
@@ -0,0 +1,181 @@
+package backend
+
+import (
+	"fmt"
+	"math"
+	"os"
+
+	"github.com/go-flac/go-flac"
+	mewflac "github.com/mewkiz/flac"
+)
+
+// AnalysisResult contains the audio analysis data
+type AnalysisResult struct {
+	FilePath      string        `json:"file_path"`
+	SampleRate    uint32        `json:"sample_rate"`
+	Channels      uint8         `json:"channels"`
+	BitsPerSample uint8         `json:"bits_per_sample"`
+	TotalSamples  uint64        `json:"total_samples"`
+	Duration      float64       `json:"duration"`
+	BitDepth      string        `json:"bit_depth"`
+	DynamicRange  float64       `json:"dynamic_range"`
+	PeakAmplitude float64       `json:"peak_amplitude"`
+	RMSLevel      float64       `json:"rms_level"`
+	Spectrum      *SpectrumData `json:"spectrum,omitempty"`
+}
+
+// AnalyzeTrack performs audio analysis on a FLAC file
+func AnalyzeTrack(filepath string) (*AnalysisResult, error) {
+	if !fileExists(filepath) {
+		return nil, fmt.Errorf("file does not exist: %s", filepath)
+	}
+
+	// Parse FLAC file
+	f, err := flac.ParseFile(filepath)
+	if err != nil {
+		return nil, fmt.Errorf("failed to parse FLAC file: %w", err)
+	}
+
+	result := &AnalysisResult{
+		FilePath: filepath,
+	}
+
+	// Extract basic audio properties from STREAMINFO block
+	if len(f.Meta) > 0 {
+		streamInfo := f.Meta[0]
+		if streamInfo.Type == flac.StreamInfo {
+			// Read STREAMINFO data
+			data := streamInfo.Data
+			if len(data) >= 18 {
+				// Sample rate (bits 10-29 of bytes 10-13)
+				result.SampleRate = uint32(data[10])<<12 | uint32(data[11])<<4 | uint32(data[12])>>4
+
+				// Channels (bits 30-32 of byte 12)
+				result.Channels = ((data[12] >> 1) & 0x07) + 1
+
+				// Bits per sample (bits 33-37 of bytes 12-13)
+				result.BitsPerSample = ((data[12]&0x01)<<4 | data[13]>>4) + 1
+
+				// Total samples (bits 38-73 of bytes 13-17)
+				result.TotalSamples = uint64(data[13]&0x0F)<<32 |
+					uint64(data[14])<<24 |
+					uint64(data[15])<<16 |
+					uint64(data[16])<<8 |
+					uint64(data[17])
+
+				// Calculate duration
+				if result.SampleRate > 0 {
+					result.Duration = float64(result.TotalSamples) / float64(result.SampleRate)
+				}
+
+				// Read min/max frame size and block size for additional analysis
+				// Min block size (bytes 0-1)
+				// Max block size (bytes 2-3)
+				// These can give us hints about encoding quality
+			}
+		}
+	}
+
+	// Analyze spectrum and calculate real audio metrics
+	spectrum, err := AnalyzeSpectrum(filepath)
+	if err != nil {
+		// Log error but continue
+		fmt.Printf("Warning: failed to analyze spectrum: %v\n", err)
+	} else {
+		result.Spectrum = spectrum
+		// Calculate dynamic range, peak, and RMS from decoded samples
+		calculateRealAudioMetrics(result, filepath)
+	}
+
+	// Set bit depth
+	result.BitDepth = fmt.Sprintf("%d-bit", result.BitsPerSample)
+
+	return result, nil
+}
+
+// calculateRealAudioMetrics calculates actual dynamic range, peak, and RMS from decoded audio
+func calculateRealAudioMetrics(result *AnalysisResult, filepath string) {
+	// Decode FLAC to get actual samples
+	samples, err := decodeFLACForMetrics(filepath)
+	if err != nil {
+		return
+	}
+
+	// Calculate peak amplitude
+	var peak float64
+	var sumSquares float64
+
+	for _, sample := range samples {
+		absVal := sample
+		if absVal < 0 {
+			absVal = -absVal
+		}
+		if absVal > peak {
+			peak = absVal
+		}
+		sumSquares += sample * sample
+	}
+
+	// Convert peak to dB (reference: 1.0 = 0 dBFS)
+	peakDB := 20.0 * math.Log10(peak)
+	result.PeakAmplitude = peakDB
+
+	// Calculate RMS (Root Mean Square)
+	rms := math.Sqrt(sumSquares / float64(len(samples)))
+	rmsDB := 20.0 * math.Log10(rms)
+	result.RMSLevel = rmsDB
+
+	// Dynamic range is the difference between peak and RMS
+	result.DynamicRange = peakDB - rmsDB
+}
+
+// decodeFLACForMetrics decodes FLAC file and returns normalized samples for metric calculation
+func decodeFLACForMetrics(filepath string) ([]float64, error) {
+	stream, err := mewflac.ParseFile(filepath)
+	if err != nil {
+		return nil, err
+	}
+	defer stream.Close()
+
+	// Limit samples to prevent memory issues (10 million samples = ~3.8 minutes at 44.1kHz)
+	maxSamples := 10000000
+	samples := make([]float64, 0, maxSamples)
+
+	// Read all audio frames
+	for {
+		frame, err := stream.ParseNext()
+		if err != nil {
+			break
+		}
+
+		// Get samples from first channel (mono or left channel)
+		var channelSamples []int32
+		if len(frame.Subframes) > 0 {
+			channelSamples = frame.Subframes[0].Samples
+		}
+
+		// Normalize samples to -1.0 to 1.0 range
+		maxVal := float64(int64(1) << (stream.Info.BitsPerSample - 1))
+		for _, sample := range channelSamples {
+			if len(samples) >= maxSamples {
+				return samples, nil
+			}
+			normalized := float64(sample) / maxVal
+			samples = append(samples, normalized)
+		}
+
+		if len(samples) >= maxSamples {
+			break
+		}
+	}
+
+	return samples, nil
+}
+
+func GetFileSize(filepath string) (int64, error) {
+	info, err := os.Stat(filepath)
+	if err != nil {
+		return 0, err
+	}
+	return info.Size(), nil
+}
@@ -48,3 +48,31 @@ func SelectFolderDialog(ctx context.Context, defaultPath string) (string, error)

 	return selectedPath, nil
 }
+
+func SelectFileDialog(ctx context.Context) (string, error) {
+	options := wailsRuntime.OpenDialogOptions{
+		Title: "Select FLAC File for Analysis",
+		Filters: []wailsRuntime.FileFilter{
+			{
+				DisplayName: "FLAC Audio Files (*.flac)",
+				Pattern:     "*.flac",
+			},
+			{
+				DisplayName: "All Files (*.*)",
+				Pattern:     "*.*",
+			},
+		},
+	}
+
+	selectedFile, err := wailsRuntime.OpenFileDialog(ctx, options)
+	if err != nil {
+		return "", err
+	}
+
+	// If user cancelled, selectedFile will be empty
+	if selectedFile == "" {
+		return "", nil
+	}
+
+	return selectedFile, nil
+}
@@ -0,0 +1,205 @@
+package backend
+
+import (
+	"fmt"
+	"math"
+	"math/cmplx"
+	"os"
+
+	"github.com/mewkiz/flac"
+)
+
+// SpectrumData contains frequency spectrum information
+type SpectrumData struct {
+	TimeSlices []TimeSlice `json:"time_slices"`
+	SampleRate int         `json:"sample_rate"`
+	FreqBins   int         `json:"freq_bins"`
+	Duration   float64     `json:"duration"`
+	MaxFreq    float64     `json:"max_freq"`
+}
+
+// TimeSlice represents spectrum data at a point in time
+type TimeSlice struct {
+	Time       float64   `json:"time"`
+	Magnitudes []float64 `json:"magnitudes"`
+}
+
+// AnalyzeSpectrum decodes FLAC file and performs FFT analysis
+func AnalyzeSpectrum(filepath string) (*SpectrumData, error) {
+	// Open FLAC file
+	stream, err := flac.ParseFile(filepath)
+	if err != nil {
+		return nil, fmt.Errorf("failed to parse FLAC: %w", err)
+	}
+	defer stream.Close()
+
+	info := stream.Info
+	sampleRate := int(info.SampleRate)
+	channels := int(info.NChannels)
+
+	// Read audio samples
+	samples, err := readSamples(stream, channels)
+	if err != nil {
+		return nil, fmt.Errorf("failed to read samples: %w", err)
+	}
+
+	if len(samples) == 0 {
+		return nil, fmt.Errorf("no audio samples found")
+	}
+
+	// Calculate spectrum
+	return calculateSpectrum(samples, sampleRate), nil
+}
+
+// readSamples reads and decodes audio samples from FLAC stream
+func readSamples(stream *flac.Stream, channels int) ([]float64, error) {
+	var allSamples []float64
+	maxSamples := 10 * 1024 * 1024 // Limit to ~10 million samples to avoid memory issues
+
+	// Decode frames
+	for {
+		frame, err := stream.ParseNext()
+		if err != nil {
+			// End of stream
+			break
+		}
+
+		// Convert samples to float64 and mix channels to mono
+		for i := 0; i < frame.Subframes[0].NSamples; i++ {
+			var sample float64
+
+			// Mix all channels to mono by averaging
+			for ch := 0; ch < channels; ch++ {
+				sample += float64(frame.Subframes[ch].Samples[i])
+			}
+			sample /= float64(channels)
+
+			allSamples = append(allSamples, sample)
+
+			// Limit sample count
+			if len(allSamples) >= maxSamples {
+				return allSamples, nil
+			}
+		}
+	}
+
+	return allSamples, nil
+}
+
+// calculateSpectrum performs FFT analysis on audio samples
+func calculateSpectrum(samples []float64, sampleRate int) *SpectrumData {
+	fftSize := 8192
+	numTimeSlices := 300
+
+	duration := float64(len(samples)) / float64(sampleRate)
+
+	samplesPerSlice := len(samples) / numTimeSlices
+	if samplesPerSlice < fftSize {
+		samplesPerSlice = fftSize
+		numTimeSlices = len(samples) / fftSize
+	}
+
+	timeSlices := make([]TimeSlice, 0, numTimeSlices)
+	freqBins := fftSize / 2
+	maxFreq := float64(sampleRate) / 2.0
+
+	for i := 0; i < numTimeSlices; i++ {
+		startIdx := i * samplesPerSlice
+		if startIdx+fftSize > len(samples) {
+			break
+		}
+
+		window := samples[startIdx : startIdx+fftSize]
+
+		windowedSamples := applyHannWindow(window)
+
+		spectrum := fft(windowedSamples)
+
+		magnitudes := make([]float64, freqBins)
+		for j := 0; j < freqBins; j++ {
+			magnitude := cmplx.Abs(spectrum[j])
+
+			if magnitude < 1e-10 {
+				magnitude = 1e-10
+			}
+			magnitudes[j] = 20 * math.Log10(magnitude)
+		}
+
+		timeSlice := TimeSlice{
+			Time:       float64(startIdx) / float64(sampleRate),
+			Magnitudes: magnitudes,
+		}
+		timeSlices = append(timeSlices, timeSlice)
+	}
+
+	return &SpectrumData{
+		TimeSlices: timeSlices,
+		SampleRate: sampleRate,
+		FreqBins:   freqBins,
+		Duration:   duration,
+		MaxFreq:    maxFreq,
+	}
+}
+
+// applyHannWindow applies Hann window to reduce spectral leakage
+func applyHannWindow(samples []float64) []float64 {
+	n := len(samples)
+	windowed := make([]float64, n)
+
+	for i := 0; i < n; i++ {
+		window := 0.5 * (1.0 - math.Cos(2.0*math.Pi*float64(i)/float64(n-1)))
+		windowed[i] = samples[i] * window
+	}
+
+	return windowed
+}
+
+// fft performs Fast Fourier Transform using Cooley-Tukey algorithm
+func fft(samples []float64) []complex128 {
+	n := len(samples)
+
+	x := make([]complex128, n)
+	for i := 0; i < n; i++ {
+		x[i] = complex(samples[i], 0)
+	}
+
+	return fftRecursive(x)
+}
+
+// fftRecursive performs recursive FFT
+func fftRecursive(x []complex128) []complex128 {
+	n := len(x)
+
+	if n <= 1 {
+		return x
+	}
+
+	even := make([]complex128, n/2)
+	odd := make([]complex128, n/2)
+
+	for i := 0; i < n/2; i++ {
+		even[i] = x[2*i]
+		odd[i] = x[2*i+1]
+	}
+
+	evenFFT := fftRecursive(even)
+	oddFFT := fftRecursive(odd)
+
+	result := make([]complex128, n)
+	for k := 0; k < n/2; k++ {
+		t := cmplx.Exp(complex(0, -2*math.Pi*float64(k)/float64(n))) * oddFFT[k]
+		result[k] = evenFFT[k] + t
+		result[k+n/2] = evenFFT[k] - t
+	}
+
+	return result
+}
+
+// GetFileSize helper
+func getSpectrumFileSize(filepath string) (int64, error) {
+	info, err := os.Stat(filepath)
+	if err != nil {
+		return 0, err
+	}
+	return info.Size(), nil
+}