The speech signals carry important information about the speaker such as gender, age, language, accent, etc. Recognition of speaker characteristics has a broad commercial, medical and forensic applications such as service customization, interactive voice response systems, natural human-machine interaction and directing the forensic investigation process.
Identifying/profiling a person based upon his/her voice is the goal for speaker characterization. Various acoustic features are being tested to find if a speaker’s identity can be detected and modeled. It mainly concentrates on finding out the features (like formant frequencies, pitch, duration, energy, etc.) which can be used to uniquely identify a speaker from the rest of the group. Standard techniques like MFCCs, Gaussian mixture models, Support vector machines, etc. are used along with a set of features to characterize a speaker. Speaker characterization has numerous applications in security, mobile devices, etc.
The work has been done in this field by Sujata Kulkarni and Abhimanyu Singh, both M Tech students.