1

keyBoard/Class/AiTalk/VM/AudioCaptureManager.m

@@ -14,6 +14,7 @@ const int kAudioChannels = 1;
 const NSUInteger kAudioFrameDuration = 20; // ms
 const NSUInteger kAudioFrameSamples = 320; // 16000 * 0.02
 const NSUInteger kAudioFrameBytes = 640;   // 320 * 2 (Int16)
+static const float kAudioSoftwareGain = 2.5f;
 
 @interface AudioCaptureManager ()
 
@@ -24,6 +25,7 @@ const NSUInteger kAudioFrameBytes = 640;   // 320 * 2 (Int16)
 // Ring buffer for accumulating samples to form 20ms frames
 @property(nonatomic, strong) NSMutableData *ringBuffer;
 @property(nonatomic, assign) NSUInteger ringBufferWriteIndex;
+@property(nonatomic, assign) NSTimeInterval lastStatsLogTime;
 
 @end
 
@@ -39,6 +41,7 @@ const NSUInteger kAudioFrameBytes = 640;   // 320 * 2 (Int16)
         initWithLength:kAudioFrameBytes * 4]; // Buffer for multiple frames
     _ringBufferWriteIndex = 0;
     _capturing = NO;
+    _lastStatsLogTime = 0;
   }
   return self;
 }
@@ -180,8 +183,11 @@ const NSUInteger kAudioFrameBytes = 640;   // 320 * 2 (Int16)
   NSUInteger sampleCount = outputBuffer.frameLength;
   NSUInteger byteCount = sampleCount * sizeof(int16_t);
 
+  [self applySoftwareGainIfNeeded:samples sampleCount:sampleCount];
+
   // 计算 RMS
   [self calculateAndReportRMS:samples sampleCount:sampleCount];
+  [self logAudioStatsIfNeeded:samples sampleCount:sampleCount];
 
   // 将数据添加到 ring buffer 并输出完整帧
   dispatch_async(self.audioQueue, ^{
@@ -266,4 +272,53 @@ const NSUInteger kAudioFrameBytes = 640;   // 320 * 2 (Int16)
   });
 }
 
+- (void)applySoftwareGainIfNeeded:(int16_t *)samples
+                      sampleCount:(NSUInteger)sampleCount {
+  if (kAudioSoftwareGain <= 1.0f || sampleCount == 0) {
+    return;
+  }
+
+  for (NSUInteger i = 0; i < sampleCount; i++) {
+    float scaled = (float)samples[i] * kAudioSoftwareGain;
+    if (scaled > 32767.0f) {
+      samples[i] = 32767;
+    } else if (scaled < -32768.0f) {
+      samples[i] = -32768;
+    } else {
+      samples[i] = (int16_t)scaled;
+    }
+  }
+}
+
+- (void)logAudioStatsIfNeeded:(int16_t *)samples
+                  sampleCount:(NSUInteger)sampleCount {
+  NSTimeInterval now = [[NSDate date] timeIntervalSince1970];
+  if (now - self.lastStatsLogTime < 1.0) {
+    return;
+  }
+  self.lastStatsLogTime = now;
+
+  if (sampleCount == 0) {
+    return;
+  }
+
+  NSUInteger nonZeroCount = 0;
+  int16_t peak = 0;
+  for (NSUInteger i = 0; i < sampleCount; i++) {
+    int16_t value = samples[i];
+    if (value != 0) {
+      nonZeroCount++;
+    }
+    int16_t absValue = (int16_t)abs(value);
+    if (absValue > peak) {
+      peak = absValue;
+    }
+  }
+
+  double nonZeroRatio = (double)nonZeroCount / (double)sampleCount;
+  double peakNormalized = (double)peak / 32768.0;
+  NSLog(@"[AudioCaptureManager] Stats: peak=%.3f nonZero=%.2f%%",
+        peakNormalized, nonZeroRatio * 100.0);
+}
+
 @end