新增: 音频可视化8种主题+节拍检测+交互效果

audio.go

@@ -74,33 +74,55 @@ func fftMagnitude(input []float32) []float64 {
 	return mag
 }
 
+// downsampleBins: 分段映射 — 低频线性保细节，中高频对数保覆盖
+// 不再做额外 tilt 补偿，映射本身就是均衡的
 func downsampleBins(magnitudes []float64, target int) []float64 {
 	srcLen := len(magnitudes)
 	if srcLen == 0 || target == 0 {
 		return make([]float64, target)
 	}
+
+	// 低频线性区间：前20%的 display bins 覆盖 FFT 1-80 (约47Hz-3.7kHz)
+	// 中高频对数区间：后80%覆盖 FFT 80-512 (3.7kHz-24kHz)
+	const lowEnd = 80 // 低频截止 FFT 索引
+	const lowFrac = 0.2
+
 	bins := make([]float64, target)
 	for i := 0; i < target; i++ {
 		frac := float64(i) / float64(target)
-		start := int(math.Pow(float64(srcLen), frac))
 		fracNext := float64(i+1) / float64(target)
-		end := int(math.Pow(float64(srcLen), fracNext))
+
+		var start, end int
+		if frac < lowFrac {
+			start = int(frac / lowFrac * float64(lowEnd))
+			end = int(fracNext / lowFrac * float64(lowEnd))
+		} else {
+			f := (frac - lowFrac) / (1 - lowFrac)
+			fNext := (fracNext - lowFrac) / (1 - lowFrac)
+			start = lowEnd + int(float64(srcLen-lowEnd)*math.Pow(f, 0.7))
+			end = lowEnd + int(float64(srcLen-lowEnd)*math.Pow(fNext, 0.7))
+		}
+
 		if end <= start {
 			end = start + 1
 		}
 		if end > srcLen {
 			end = srcLen
 		}
-		sum := 0.0
+
+		// 取最大值保留峰值冲击力
+		maxVal := 0.0
 		for j := start; j < end; j++ {
-			sum += magnitudes[j]
+			if magnitudes[j] > maxVal {
+				maxVal = magnitudes[j]
+			}
 		}
-		bins[i] = sum / float64(end-start)
+		bins[i] = maxVal
 	}
 	return bins
 }
 
-func smoothBins(prev, curr []float64, factor float64) []float64 {
+func smoothBinsGo(prev, curr []float64, factor float64) []float64 {
 	n := len(curr)
 	if len(prev) != n {
 		return curr
@@ -122,7 +144,6 @@ func comCall(obj uintptr, vtableIdx uintptr, args ...uintptr) uintptr {
 	}
 	vtable := *(*uintptr)(unsafe.Pointer(obj))
 	method := *(*uintptr)(unsafe.Pointer(vtable + vtableIdx*unsafe.Sizeof(uintptr(0))))
-	// 构建参数数组 (this + args)，最多支持 12 参数
 	a := make([]uintptr, 12)
 	a[0] = obj
 	for i, arg := range args {
@@ -155,27 +176,23 @@ var (
 	iidIAudioCaptureClient   = windows.GUID{Data1: 0xC8ADBD64, Data2: 0xE71E, Data3: 0x48A0, Data4: [8]byte{0xA4, 0xDE, 0x18, 0x5C, 0x39, 0x5C, 0xD3, 0x17}}
 )
 
-// IMMDeviceEnumerator vtable (IUnknown 3 + 自身方法)
 const (
-	enumeratorGetDefaultAudioEndpoint = 4 // 第 4 个方法 (0-based from vtable start)
+	enumeratorGetDefaultAudioEndpoint = 4
 )
 
-// IMMDevice vtable
 const (
-	deviceActivate = 3 // 第 3 个方法
+	deviceActivate = 3
 )
 
-// IAudioClient vtable
 const (
-	audioClientInitialize   = 3
+	audioClientInitialize    = 3
 	audioClientGetBufferSize = 4
-	audioClientGetMixFormat  = 8
-	audioClientStart         = 10
-	audioClientStop          = 11
-	audioClientGetService    = 14
+	audioClientGetMixFormat   = 8
+	audioClientStart          = 10
+	audioClientStop           = 11
+	audioClientGetService     = 14
 )
 
-// IAudioCaptureClient vtable
 const (
 	captureGetBuffer     = 3
 	captureReleaseBuffer = 4
@@ -191,6 +208,7 @@ var (
 	audioDone   chan struct{}
 	audioActive bool
 	smoothed    []float64
+	runningMax  float64
 )
 
 func startAudioCapture() {
@@ -203,6 +221,7 @@ func startAudioCapture() {
 	audioDone = make(chan struct{})
 	audioActive = true
 	smoothed = nil
+	runningMax = 0
 	go audioCaptureLoop()
 	log.Println("音频捕获: 已启动")
 }
@@ -236,15 +255,13 @@ func audioCaptureLoop() {
 	runtime.LockOSThread()
 	defer runtime.UnlockOSThread()
 
-	// COM 初始化
 	ole32dll.NewProc("CoInitializeEx").Call(0, 0)
 	defer ole32dll.NewProc("CoUninitialize").Call()
 
-	// CoCreateInstance -> IMMDeviceEnumerator
 	var enumerator uintptr
 	hr, _, _ := ole32dll.NewProc("CoCreateInstance").Call(
 		uintptr(unsafe.Pointer(&clsidMMDeviceEnumerator)),
-		0, 1, // CLSCTX_INPROC_SERVER
+		0, 1,
 		uintptr(unsafe.Pointer(&iidIMMDeviceEnumerator)),
 		uintptr(unsafe.Pointer(&enumerator)),
 	)
@@ -255,7 +272,6 @@ func audioCaptureLoop() {
 	}
 	defer comRelease(enumerator)
 
-	// enumerator->GetDefaultAudioEndpoint(eRender=0, eConsole=0)
 	var device uintptr
 	hr = comCall(enumerator, enumeratorGetDefaultAudioEndpoint, 0, 0, uintptr(unsafe.Pointer(&device)))
 	if hr != 0 || device == 0 {
@@ -265,7 +281,6 @@ func audioCaptureLoop() {
 	}
 	defer comRelease(device)
 
-	// device->Activate(IID_IAudioClient, CLSCTX_INPROC=1, NULL, &audioClient)
 	var audioClient uintptr
 	hr = comCall(device, deviceActivate,
 		uintptr(unsafe.Pointer(&iidIAudioClient)),
@@ -279,7 +294,6 @@ func audioCaptureLoop() {
 	}
 	defer comRelease(audioClient)
 
-	// audioClient->GetMixFormat(&pwfx)
 	var pwfx uintptr
 	hr = comCall(audioClient, audioClientGetMixFormat, uintptr(unsafe.Pointer(&pwfx)))
 	if hr != 0 || pwfx == 0 {
@@ -293,11 +307,10 @@ func audioCaptureLoop() {
 	nSamplesPerSec := *(*uint32)(unsafe.Pointer(pwfx + 4))
 	log.Printf("音频捕获: 格式 ch=%d rate=%d", nChannels, nSamplesPerSec)
 
-	// audioClient->Initialize(SHARED, LOOPBACK, 1s buffer, 0, pwfx, NULL)
 	hr = comCall(audioClient, audioClientInitialize,
-		0, // AUDCLNT_SHAREMODE_SHARED
-		0x00020000, // AUDCLNT_STREAMFLAGS_LOOPBACK
-		10000000,   // 1 second (100ns units)
+		0,
+		0x00020000,
+		10000000,
 		0,
 		pwfx,
 		0,
@@ -308,7 +321,6 @@ func audioCaptureLoop() {
 		return
 	}
 
-	// audioClient->GetService(IID_IAudioCaptureClient, &captureClient)
 	var captureClient uintptr
 	hr = comCall(audioClient, audioClientGetService,
 		uintptr(unsafe.Pointer(&iidIAudioCaptureClient)),
@@ -321,11 +333,9 @@ func audioCaptureLoop() {
 	}
 	defer comRelease(captureClient)
 
-	// audioClient->Start()
 	comCall(audioClient, audioClientStart)
 	defer comCall(audioClient, audioClientStop)
 
-	// 环形缓冲区 (~100ms)
 	ch := int(nChannels)
 	bufSize := int(nSamplesPerSec)/10*ch + fftSize*ch
 	ringBuf := make([]float32, bufSize)
@@ -337,6 +347,9 @@ func audioCaptureLoop() {
 		smoothingFactor = 0.7
 	}
 
+	// 预分配 mono slice
+	mono := make([]float32, fftSize)
+
 	ticker := time.NewTicker(33 * time.Millisecond)
 	defer ticker.Stop()
 
@@ -348,7 +361,6 @@ func audioCaptureLoop() {
 		case <-ticker.C:
 		}
 
-		// 读取 WASAPI 缓冲区
 		for {
 			var pData, nFrames, flags uintptr
 			hr = comCall(captureClient, captureGetBuffer,
@@ -368,8 +380,6 @@ func audioCaptureLoop() {
 			comCall(captureClient, captureReleaseBuffer, uintptr(nFrames))
 		}
 
-		// 取 fftSize 个单声道样本
-		mono := make([]float32, fftSize)
 		for i := 0; i < fftSize; i++ {
 			pos := (writePos - fftSize*ch + i*ch + bufSize) % bufSize
 			var sum float32
@@ -383,20 +393,25 @@ func audioCaptureLoop() {
 		bins := downsampleBins(mag, targetBins)
 
 		if smoothed != nil {
-			bins = smoothBins(smoothed, bins, smoothingFactor)
+			bins = smoothBinsGo(smoothed, bins, smoothingFactor)
 		}
 		smoothed = bins
 
-		// 归一化
-		maxVal := 0.0
+		// 全局归一化（慢衰减 runningMax）
+		curMax := 0.0
 		for _, v := range bins {
-			if v > maxVal {
-				maxVal = v
+			if v > curMax {
+				curMax = v
 			}
 		}
-		if maxVal > 0.001 {
+		if curMax > runningMax {
+			runningMax = curMax
+		} else {
+			runningMax *= 0.985
+		}
+		if runningMax > 0.001 {
 			for i := range bins {
-				bins[i] /= maxVal
+				bins[i] /= runningMax
 			}
 		}
 
@@ -407,6 +422,6 @@ func audioCaptureLoop() {
 
 func comRelease(obj uintptr) {
 	if obj != 0 {
-		comCall(obj, 2) // IUnknown::Release = vtable index 2
+		comCall(obj, 2)
 	}
 }