Table 2 shows the classification results with the number of HMM states equal to 4. The first column represents different
number of mixtures while the first row represents different analysis frame size, i.e., window length with fixed window shifting rate. By observing the results of each column, we can find that the best performance is achieved when the number of mixtures is set to 8. In the light of this observation, we can say that neither too small nor too large number of mixtures could produce the best results because too small number of mixtures couldn’t model the spectral variability of heart sounds well in each HMM state while too big number would result in mixture-heavies problem. Similarly, we can observe the results in each row to investigate the influence of the window length on the classification performance. Firstly by observing the results of the first row and the fourth row, we can find that the performance decreases as the window length gets smaller.
Table 2 shows the classification results with the number of HMM states equal to 4. The first column represents differentnumber of mixtures while the first row represents different analysis frame size, i.e., window length with fixed window shifting rate. By observing the results of each column, we can find that the best performance is achieved when the number of mixtures is set to 8. In the light of this observation, we can say that neither too small nor too large number of mixtures could produce the best results because too small number of mixtures couldn’t model the spectral variability of heart sounds well in each HMM state while too big number would result in mixture-heavies problem. Similarly, we can observe the results in each row to investigate the influence of the window length on the classification performance. Firstly by observing the results of the first row and the fourth row, we can find that the performance decreases as the window length gets smaller.
การแปล กรุณารอสักครู่..