Update sonogram_utility.py

sonogram_utility.py

@@ -123,22 +123,24 @@ def splitIntoTimeSegments(testFile,maxDurationInSeconds=60):
 def audioNormalize(waveform,sampleRate,stepSizeInSeconds = 2,dbThreshold = -50,dbTarget = -5):
     print("In audioNormalize")
     copyWaveform = waveform.clone().detach()
-    copyWaveform_db = waveform.clone().detach()
-    print("Copies made")
+    print("Waveform copy made")
     transform = torchaudio.transforms.AmplitudeToDB(stype="amplitude", top_db=80)
-    copyWaveform_db = transform(copyWaveform_db)
-    print("DB levels calculated")
     currStart = 0
     currEnd = int(min(currStart + stepSizeInSeconds * sampleRate, len(copyWaveform_db[0])-1))
     done = False
     while(not done):
-        if torch.max(copyWaveform_db[0][currStart:currEnd]).item() > dbThreshold:
-            gain = torch.min(dbTarget - copyWaveform_db[0][currStart:currEnd])
+        copyWaveform_db = waveform[:,currStart:currEnd].clone().detach()
+        copyWaveform_db = transform(copyWaveform_db)
+        if currStart == 0:
+            print("First DB level calculated")
+        
+        if torch.max(copyWaveform_db[0]).item() > dbThreshold:
+            gain = torch.min(dbTarget - copyWaveform_db[0])
             adjustGain = torchaudio.transforms.Vol(gain,'db')
             copyWaveform[0][currStart:currEnd] = adjustGain(copyWaveform[0][currStart:currEnd])
         if len(copyWaveform_db) > 1:
-            if torch.max(copyWaveform_db[1][currStart:currEnd]).item() > dbThreshold:
-                gain = torch.min(dbTarget - copyWaveform_db[1][currStart:currEnd])
+            if torch.max(copyWaveform_db[1]).item() > dbThreshold:
+                gain = torch.min(dbTarget - copyWaveform_db[1])
                 adjustGain = torchaudio.transforms.Vol(gain,'db')
                 copyWaveform[1][currStart:currEnd] = adjustGain(copyWaveform[1][currStart:currEnd])
         currStart += int(stepSizeInSeconds * sampleRate)