The vocal tract as a biometric: output measures, interrelationships, and efficacy

This paper explores methods for characterising individual voices using different vocal tract output measures. Mel frequency cepstral coefficients (MFCCs), long-term formant distributions (LTFDs) and scores based on vocal profile analysis (VPA) of long-term supralaryngeal settings were extracted from the same corpus of recordings. Distances between speakers were calculated and used to test the interrelationships between the three output measures. Strong correlations were found between the MFCC and LTFD distances, while considerably weaker correlations were found between the acoustic measures and the VPA-based distances. This suggests that while the two measures of acoustic output provide similar information, the auditory VPA offers different information relevant for voice characterisation. In a forensic context this finding is important since it suggests that it may be possible to complement acoustic analysis with VPA to improve system performance.