Note: For all experiments, the training and gallery sets are the same.

MIT AI Data Experiment #2 A. Day 030101 vs. days 031601, 032201, and 051001 Probe sequences Gallery sequences Result (HMM Approach) Similarity matrix (image) Similarity matrix (.mat) Cumulative Match Score Similarity files Probe sequences Gallery sequences Result (Algorithm II) Similarity matrix (image) Similarity matrix (.mat) Cumulative Match Score Similarity files B. Day 031601 vs. days 030101, 032201, and 051001 Probe sequences Gallery sequences Result (HMM Approach) Similarity matrix (image) Similarity matrix (.mat) Cumulative Match Score Similarity files Probe sequences Gallery sequences Result (Algorithm II) Similarity matrix (image) Similarity matrix (.mat) Cumulative Match Score Similarity files C. Day 032201 vs. days 030101, 031601, and 051001 Probe sequences Gallery sequences Result (HMM Approach) Similarity matrix (image) Similarity matrix (.mat) Cumulative Match Score Similarity files Probe sequences Gallery sequences Result (Algorithm II) Similarity matrix (image) Similarity matrix (.mat) Cumulative Match Score Similarity files D. Day 051001 vs. days 030101, 031601, and 032201 Probe sequences Gallery sequences Result (HMM Approach) Similarity matrix (image) Similarity matrix (.mat) Cumulative Match Score Similarity files Probe sequences Gallery sequences Result (Algorithm II) Similarity matrix (image) Similarity matrix (.mat) Cumulative Match Score Similarity files A note on the results using HMM approach An element essential to our training/testing philosophy is the HMM model that we build for every subject in the database. With every additional labeled incoming sequence during training, we update the model for that particular subject. Therefore, the number of models built at the end of a training procedure is equal to the of subjects present in that particular experiment. To find a common ground between the experiments designed by MIT and our algorithm, we have reorganized the MIT dataset as follows: MIT convention:         day --> sequence Modified convention:  day --> subject --> sequences of the subject And performed the experiments as designed by MIT (training with data from 3 days and testing with the 4th).      Site First Created: 8/3/01 | Last Modified: 04/11/2002