Pitch period estimation base on voiced degree weighted sub-frame octave region dynamic programming

Pitch period is one of the most important parameters of speech signal. Different form traditional dynamic programming (DP) methods, voiced degree weighted sub-frame octave region parameter DP algorithm is proposed. Judge transition sound by sub-frame energy variation to diminish pitch hopping. Two groups of parameters were extracted within each frame respectively, and the primary pitch and octave region parameter were confirmed as candidates, then DP method in the chosen octave region between sub-frames was executed with voiced degree factor introduced. Experiments based on standard broadcast speech show that algorithm proposed reduces double/semi-double error effectively. The gross error reduced by 3.45% and mean normal error reduced by 35.24%, compared with traditional way.