It is well known that speaker identification performs extremely well in theneutral talking environments; however, the identification performance isdeclined sharply in the shouted talking environments. This work aims atproposing, implementing and testing a new approach to enhance the declinedperformance in the shouted talking environments. The new proposed approach isbased on gender-dependent speaker identification using Suprasegmental HiddenMarkov Models (SPHMMs) as classifiers. This proposed approach has been testedon two different and separate speech databases: our collected database and theSpeech Under Simulated and Actual Stress (SUSAS) database. The results of thiswork show that gender-dependent speaker identification based on SPHMMsoutperforms gender-independent speaker identification based on the same modelsand gender-dependent speaker identification based on Hidden Markov Models(HMMs) by about 6% and 8%, respectively. The results obtained based on theproposed approach are close to those obtained in subjective evaluation by humanjudges.
展开▼
