| The speaker gender recognition and age estimation can be accomplished through analysisand processing of the speaker’s voic. To be able to tell a person’s gender and age simply byhis voice is of important significance. For example, with the knowledge of the user’s genderand age, a human-computer interaction system can determine the specific user group that theuser belongs to, so that it can interact with the user more accordingly. Moreover, authorizationcan also be carried out according to the gender and age of the user. In criminal investigation,knowledge of the speaker’s gender and age also helps to narrow the scope of criminalinvestigation. It can also provide basis for criminal psychology analysis and behavior analysis.As pre-classification techniques, the speaker gender recognition and age estimation are ofimportant significance to the development of speaker recognition. They can accelerate thespeed of operation, improve the recognition rate of speaker recognition and alsoprovide decision basis for speaker recognition.This thesis thoroughly analyzes voiceprint characteristics that are associated with genderand age of the speaker. The main accomplishments of this thesis are as follows.1.An experimental speech database according to the gender and age is established.The speakers are divided into five different age groups, and then classified according to theirgender. Ten kinds of speaker database are built up from scratch.2.On the basis of the analysis of the distinctive voice characteristics that are related to thegender and age of the speaker, the more distinctive MFCC and LPCC are selected as the mainfeature for gender recognition and age estimation.3.This thesis presents an implementation of the MFCC&GMM based speaker genderrecognition system. It also presents an implementation of the MFCC-GMM and LPCC basedspeaker gender and age estimation system.4.The fourth one is the analysis of the role that gender recognition and age estimationplay in a speaker recognition system. |
PDF Full Text Request