tuned algorithm and renamed langlang to london

.gitignore

@@ -1,4 +1,5 @@
 .DS_Store
+demo
 
 # Node
 logs

Compare.py

@@ -3,7 +3,7 @@
 import matplotlib.pyplot as plt
 import matplotlib.patches as mpatches
 
-filenames = ["columbia", "philadelphia", "royal", "slovak", "langlang"]
+filenames = ["columbia", "philadelphia", "royal", "slovak", "london"]
 
 data = {}
 for filename in filenames:

Generate_Frequencies.py

@@ -53,23 +53,26 @@
 # first, identify all the outliers and move them closer to the correct position
 corrected = [pitches[0]]
 for i in range(1, len(pitches)-1):
-    diff = abs(pitches[i]-corrected[i-1])
-    if diff > 150:
+    diff = abs(pitches[i]-corrected[i-1]) / corrected[i-1]
+    if diff > 0.4:
         corrected += [corrected[i-1]]
-    elif diff > 100:
+        #print(1, i, pitches[i], corrected[i-1], diff, average)        
+    elif diff > 0.25:
         average = (corrected[i-1] + pitches[i+1]) / 2
+        #print(1, i, pitches[i], corrected[i-1], diff, average)
         corrected += [average]
     else:
         corrected += [pitches[i]]
-
+        
 # then iterate over the corrected array, gradually bringing the outliers into line
 for x in range(2,20):
     arr = corrected
     for i in range(1, len(arr)-1):
-        diff = abs(arr[i] - arr[i+1])
-        threshold =100 - 5 * x
+        diff = abs(arr[i] - arr[i+1]) / arr[i+1]
+        threshold = 1 - 0.06 * x
         if diff > threshold:
             average = (arr[i-1] + arr[i+1]) / 2
+            #print(x, i, threshold, arr[i], corrected[i-1], diff, average)
             corrected[i] = average
 
 fig = plt.figure(figsize=(12, 9), frameon=True)

README.md

@@ -17,7 +17,7 @@ You'll need Node and Python 3, which you can download directly or install via Ho
 + [Philadelphia Orchestra](https://www.youtube.com/watch?v=xWB5m3ycYg0)
 + [Royal Philharmonic Orchestra](https://www.youtube.com/watch?v=hpynmrUI4oI)
 + [Slovak National Philharmonic Orchestra](https://www.youtube.com/watch?v=ynEOo28lsbc)
-+ [Lang Lang](https://www.youtube.com/watch?v=ss2GFGMu198)
++ [London Symphony Orchestra](https://www.youtube.com/watch?v=ss2GFGMu198)
 
 ## Getting the clips
 
@@ -27,7 +27,7 @@ The [`get_audio.js`](get_audio.js) script downloads the YouTube url, extracts ju
 + `node get_audio.js --name=philadelphia --url=https://www.youtube.com/watch?v=xWB5m3ycYg0 --start=3 --end=9`
 + `node get_audio.js --name=royal --url=https://www.youtube.com/watch?v=hpynmrUI4oI --start=3 --end=8`
 + `node get_audio.js --name=slovak --url=https://www.youtube.com/watch?v=ynEOo28lsbc --start=5 --end=8`
-+ `node get_audio.js --name=langlang --url=https://www.youtube.com/watch?v=ss2GFGMu198 --start=6 --end=12`
++ `node get_audio.js --name=london --url=https://www.youtube.com/watch?v=ss2GFGMu198 --start=6 --end=12`
 
 (I cleaned up each sample in Audacity to make sure it starts right at the beginning of the glissando and ends right before the rest of the orchestra joins in)
 
@@ -75,7 +75,7 @@ Here are the command-line scripts for all five samples. (Again, depending on you
 + `python Generate_Frequencies.py --name=philadelphia --title="Philadelphia Orchestra"`
 + `python Generate_Frequencies.py --name=royal --title="Royal Philharmonic Orchestra"`
 + `python Generate_Frequencies.py --name=slovak --title="Slovak National Philharmonic Orchestra"`
-+ `python Generate_Frequencies.py --name=langlang --title="Lang Lang"`
++ `python Generate_Frequencies.py --name=london --title="London Symphony Orchestra"`
 
 Let's compare them and see how they stack up!