A recognition method with parametric trajectory synthesized using direct relations between static and dynamic feature vector time series.
Yasuhiro MinamiErik McDermottAtsushi NakamuraShigeru KatagiriPublished in: ICASSP (2002)
Keyphrases
- computational cost
- feature vectors
- feature set
- detection method
- multivariate time series
- significant improvement
- cost function
- pairwise
- similarity measure
- recognition rate
- recognition accuracy
- classification method
- euclidean distance
- segmentation method
- speech recognition
- distance function
- clustering method
- dynamic programming
- hidden markov models
- pattern recognition