Between-speaker variability in dynamic formant characteristics in spontaneous speech

Responsible: Dr. Lei He
Duration of the project: June 1 2019 – May 31 2020

Project Description:

Source signals, vocal tract resonances, and articulatory movements are the essential processes of speech production. Each of these processes encodes speaker-specific information. Quantifying individual differences in these processes thus is essential in forensic phonetic practices. Measuring speaker-specific characteristics in both glottal vibrations and vocal tract resonances have been extensively studied in forensic phonetics and automatic speaker recognition. Nonetheless, far less attention has been paid to the temporal characteristics of speech articulation, simply because directly characterizing speaker-specific articulatory movements is almost impossible, as kinematic data of articulators are rare in either criminal or suspect materials.
To crack this conundrum, forensic speech scientists focus on acoustic properties in the speech signal that are – although not entirely – modulated by the articulatory movements. For example, Dellwo and colleagues measured speech rhythm in terms of the durational variability of various phonetic intervals (e.g., Dellwo et al. 2015, Leemann et al. 2014) or syllabic intensity variability (e.g., He and Dellwo 2014, 2016); McDougall (2006) approached the formant trajectories using least-squares polynomial approximations; and He and Dellwo (2017) measured the dynamic characteristics of the intensity contours. More recently, He et al. (2019) combined the ideas of both McDougall (2006) and He and Dellwo (2017) and measured the dynamic characteristics of the first formant, and found that the speeds of F1 decreases contained more speaker-specific information than speeds of F1 increases. However, this study (He et al. 2019) only focused on read speech in laboratory settings. To evaluate the practical value of this method in forensic practices, it is important to test whether the same results will be obtained using spontaneous speech, in different languages, which is the aim of this proposed project.

Publications (or conference presentation):

He, Lei; Zhang, Yu; Dellwo, Volker (2019). Between-speaker variability and temporal organization of the first formant. Journal of the Acoustical Society of America 145(3), pp. EL209–EL214. 

Keywords: between-speaker variability, formant dynamics, spontaneous speech

Funding source: International Association for Forensic Phonetics and Acoustics

Partner: Dr. Willemijn Heeren, Universiteit Leiden, the Netherlands