Speaker Recognition Robust for Time Difference based on Sub-space Method
Yasuo Ariki · Institutional Repositories DataBase (IRDB) · 1994
This paper proposes a method to separate speaker information from phonetic information included in speech data. A speaker recognition method using the separated speaker information is also proposed and shown to be equivalent with a method based on speaker subspace. A new speaker recognition method robust for time session difference is also proposed. The validity of this proposed method was verified by carring out simple speaker recognition experiments.