Speech Signal Processing: Homework-3             Fall, 2003

 

 

Train two three-state Hidden Markov Models (HMMs) using the following labeled training
sequences.

文字方塊: The Prototype HMM

 

 

Class 1:

1.  ABBCABCAABC  

2.  ABCABC            

3.  ABCAABC          

4.  BBABCAB           

5.  BCAABCCAB      

6.  CACCABCA        

7.  CABCABCA        

8.  CABCA              
9.  BCABCA

 

 

Class 2:

1.  BBBCCBC   

2.  CCBABB     

3.  AACCBBB   

4.  BBABBAC   

5.  CCAABBAB 

6.  BBBCCBAA 

7.  ABBBBABA 

8.  CCCCC       
9.  BBAAA

        

 

　

P1. Please specify the model parameters after the first and 50th iterations of Baum-Welch training.

P2. Please show the recognition results by using the above training sequences as the testing
       data (The so-called inside testing). You have to perform the recognition task with the HMMs trained from
       the first and 50th iterations of Baum-Welch training, respectively.

P3. Which class do the following testing sequences belong to?
              ABCABCCAB
              AABABCCCCBBB

　

P4. What are the results if Observable Markov Models were instead used  in P1, P2 and P3?