Speech recognition Spring 2018, NYUST
Using DTW on the MFCC's to match a spoken word to a template
- Use MFCC to analysis audio's features
- Use DTW to find the shortest path
• If the audio which we want to compare have different length(vector), we use DTW to find the distance between them.
There are three parts of DTW:
• Determin D(i , j) = Distance between ti, rj;
• A(i, j) = the shortest cumulative distance value (begin point(1, 1) to end point (i, j))
• Use Reccurence Relation to find the shortest cumulative distance value between vector(t) and vector(r)