machine learning - Vector Quantization Algorithms used to provide observation sequences for Hidden Markov Models -

i building gesture recognition application. extracting features hand (such [angle of motion, width: length ratio,...]). feature vector. obviously, 1 cannot have observation of such vectors input hidden markov model.

general info: gesture made of set of postures.

thus found paper suggests vector quantization. aim use sort of algorithm feed feature vector 1 side, , obtain integer or simple value (which maps particular state/posture). feed these set of symbols (mapping complete gesture) hidden markov model (

one algorithm k means clustering. unfortunately, if example recognize x amount of symbols, , decided go creating cluster each symbol, k means creates maximum of x clusters, cannot used map each state/posture have.

i thinking need sort of clustering algorithm, if possible, can supervised dictate these sets of feature vectors map posture , others map posture b.

does there exist form of supervised clustering/ classification algorithm input set of vectors [angle, width, height] , obtain simple symbol e.g. 'a'?

the following sample data generated:

posture a: [angle of ellipse surrounding it, height:width ratio]

• 0.802985 33.909615
• 0.722824 31.209663
• 0.734535 30.206722
• 0.68397 31.838253
• 0.713706 34.29641
• 0.688798 30.603661
• 0.721395 34.880161

posture b: [structured same posture a]

• 0.474164 16.077467
  • 0.483104 14.526289
  • 0.478904 14.800572
  • 0.483134 14.523611
  • 0.480608 14.41159
  • 0.481552 15.563665
  • 0.497951 15.563585