A Study Of Speaker Comparison Based On Formants And Information Elements

The acoustic-phonetic analysis is the main method used for speaker comparison at home and abroad currently.However,the accuracy of this method is seriously subjected to many conditions such as the presence of background noise,differences in transmission channels and recording equipments,and capability of the appraiser.Under this condition,the present study attempts to make comparison by conducting both the acoustic-phonetic analysis and discourse information analysis for speech samples,hoping to offer more theoretical support for the validity of the appraisal conclusion provided by phonetic experts in court.Both quantitative and qualitative approaches are adopted to conduct this study.Quantitative approach is taken to analyze formant parameters in the acoustic-phonetic analysis.As for discourse information analysis,qualitative method is conducted initially for the purpose of discourse information feature extracting.Quantitatively,the present research tests the discourse information elements mainly through distribution test of SPSS for their distinguishing power.It is found that F1,F2,F3,and F4 all show a certain degree of speaker distinguishing ability in real life speeches,among which F4 displays the best performance.Furthermore,all the three kinds of information elements,i.e.process,entity and condition are very useful to distinguish the questioned speech samples,especially process and entity.The present study combines the acoustic-phonetic analysis with a brand new one –DIT approach to identify the questioned speech sample so as to draw more convincing conclusions.In spite of some limitations,this study does provide a new perspective for speaker comparison and broadens the application of the discourse information analysis as well.