So I changed my algorithm so that HMM can accept different size of sequence.
(currently, I normalize the size and then calculates log-likelihood)
After that, I could get reasonable result.
You can download graph here.
| Normal HMM (All) | Weight HMM (All) | Weight (Andrew) | Weight (Jin) | Weight (Scott) |
Task1 | 81.5394 | 87.9466 | 101.2397 | 84.4787 | 81.2795 |
Task2 | 82.3386 | 95.0149 | 95.3747 | 84.4096 | 105.0517 |
Task3 | 79.2707 | 92.4030 | 109.0333 | 88.4455 | 83.9879 |
Task4 | 87.8758 | 88.6020 | 92.6539 | 91.2206 | 84.0128 |
This is the result from HMM. (small number means good motion - close to trained motion)
As you can see, almost all weighted motions are higher than normal motions.
Scott's task4's score is better when he put on weight band. And it made normal HMM score and weight HMM score similar. Except for this, we can conclude that HMM seems reflect motion's quality well.
-Jin
Normal:
Ex: Energy
--------------
1 168.53
2 66.442
3 94.084
4 483.74
Weighted:
Ex: Energy
--------------
1 221.3
2 57.46
3 56.263
4 205.54
Ex. Avg. Normal Avg. With Weights
------------------------------------
1 2079.35 4877.96
2 5925.08 3834.49
3 20313.27 17464.42
4 15911.03 18382.09
Ex. With Weights (min,mean,max,(stdev))
---------------------
1 1654.56, 2404.40, 3048.16 (438.37)
2 378.62, 957.23, 2641.28, (657.71)
3 17518.23, 26805.36, 41087.51 (5961.64)
4 7375.40, 17831.87, 32354.52 (8689.43)
Ex. With Weights
---------------------
1 1731.9, 3284.3, 4548.3 (771.31)
2 3947.8, 5788.8, 9190.8 (1820.9)
3 4298, 10568, 20295 (4663.9)
4 4806.2, 9506, 14470 (3355.6)
Ex. With Weights
---------------------
1 2464.1, 8945.2, 20658 (6588.5)
2 2658.5, 4757.4, 8741.8 (1936)
3 8992.3, 15020, 19028 (3275.4)
4 14906, 27808, 46056 (10216)