Tesis doctoralInfoling 4.31 (2015)

Autor/a: San Segundo Fernández, Eugenia
Fecha de lectura o defensa: 15 de diciembre de 2014
Título de la tesis: Forensic speaker comparison of Spanish twins and non-twin siblings. A phonetic-acoustic analysis of formant trajectories in vocalic sequences, glottal source parameters and cepstral characteristics
Director/a de tesis: Joaquim Llisterri Boix
Codirector/a: Juana Gil Fernández
Universidad: Consejo Superior de Investigaciones Científicas, Universidad Internacional Menéndez Pelayo
Departamento: Laboratorio de Fonética
País: España
Descripción de la tesis

The objective of this thesis is to investigate the phonetic-acoustic similarities and differences in three main speaker groups: monozygotic (MZ) twins, dizygotic (DZ) twins and non-twin siblings. From a forensic-phonetic perspective, the study of this type of speakers is highly relevant, as they represent extreme examples of physical similarity. Distinguishing their voices poses a well-recognized challenge in the forensic realm [see Chapter 2]. Yet, there is an interest in this investigation per se, as the study of genetically identical speakers (MZ twins) and their comparison with non-genetically-identical siblings (DZ twins and non-twin siblings), on the one hand, and with a reference population of unrelated speakers, on the other hand, allows gaining insight into the contribution of nurture and nature in the speech patterns of speakers in general. In other words: to what extent is our voice determined by our DNA and to what extent is it due to educational influences?

Besides, this study could be considered the first investigation into the phonetic and acoustic characteristics of Spanish-speaking twins and siblings. According to our research objective, a three-folded approach has been undertaken, as will be described below. For the 54 male speakers recorded ad hoc for this study, three different analyses have been carried out. On the one hand, we have labeled and analyzed the F1-F3 formant trajectories of 19 Spanish vocalic sequences. Secondly, several naturally sustained [e:] tokens have been extracted from the speakers’ spontaneous vowel fillers and their glottal source characteristics have been analyzed. These two approaches have been complemented with an automatic speaker recognition analysis carried out with the software Batvox.


Área temática: Fonética, Lingüística forense, Sociolingüística


Índice

1. INTRODUCTION
1.1. Introduction to the investigation
1.2. Defining Forensic Phonetics: current methodologies in Forensic Speaker Comparison
1.2.1. Towards a definition of Forensic Phonetics: terminology controversies
1.2.2. Current methodologies and most frequently analyzed parameters in Forensic Speaker Comparison
1.3. Methodological approach of this investigation
1.4. Outline of research objectives

2. REVIEW OF THE LITERATURE: TWIN STUDIES
2.1. Introduction
2.2. The biological bases of twinning and the twin method
2.3. Forensic relevance of twins’ voices
2.4. Phonetic studies on twins
2.4.1. Perceptual studies
2.4.2. Acoustical perspectives
2.4.3. Articulatory studies
2.4.4. Automatic approaches

3. METHOD
3.1. Introduction
3.2. Participants
3.3. Corpus design
3.3.1. First task: semi-structured spontaneous conversation
3.3.2. Second task: Fax exchange to elicit specific vocalic sequences
3.3.3. Third task: Reading of two phonetically-balanced texts
3.3.4. Fourth task: Mathematical calculations
3.3.5. Fifth task: Informal interview with the researcher
3.3.6. Vocal Control Techniques
Maximum Phonation Time
The s/a Ratio
3.4. Recording procedure
3.4.1. Materials and technical characteristics of the recording
3.4.2. Data collection set-up
3.5. Telephone filtering procedure
3.6. The likelihood-ratio approach

4. ANALYSIS OF FORMANT TRAJECTORIES
4.1. Objectives and justification
4.1.1. Objectives
4.1.2. Justification
4.2. Speech material, analysis tools and method
4.2.1. Speech material
4.2.2. Analysis tools and method
Acoustic analysis
Curve fitting
Likelihood ratio calculation
Goodness of fit (for the parametric curves) and fusion techniques
4.3. Parameters
4.4. Results
4.4.1. Curve fitting: best correlation values
4.4.2. Combination/Fusion techniques: comparing MZ, DZ, B and US tests
4.4.3. Accuracy assessment
4.5. Discussion
4.5.1. Curve fitting: best correlation values
4.5.2. Combination/Fusion techniques: comparing MZ, DZ, B and US tests
4.5.3. Accuracy assessment
4.6. Conclusions

5. GLOTTAL SOURCE ANALYSIS
5.1. Objectives and justification
5.1.1. Objectives
5.1.2. Justification
5.2. Speech material, analysis tools and method
5.2.1. Speech material
Speech material extraction
5.2.2. Analysis tools and method
5.3. Parameters
5.4. Results
5.4.1. Pilot experiment (proof of concept)
Overall results
Diagnosis of specific cases (unexpected results)
5.4.2. Final voice analysis and forensic comparison
Parameter Discrimination Capability
5.5. Discussion
5.5.1. Discussion for the first hypothesis
Intra-speaker results
Inter-speaker results
5.5.2. Discussion for the second hypothesis
5.6. Conclusions

6. AUTOMATIC ANALYSIS
6.1. Objectives and justification
6.1.1. Objectives
6.1.2. Justification
6.2. Speech material, analysis tools and method
6.2.1. Speech material
6.2.2. Analysis tools and method
6.3. Parameters
6.4. Results
6.4.1. Overall system performance
6.4.2. Sibling effect
6.4.3. Special case study: MZ twins
6.5. Discussion
6.6. Conclusions

7. CONCLUSIONS
7.1. Summary of the research approach and main conclusions
7.2. Original contributions to the research field
7.3. Implications
7.4. Limitations of the study and directions for future research

REFERENCES

APPENDIX A: Questionnaires
A1: Online questionnaire
A2: Participant questionnaire for the first recording session
A3: Participant questionnaire for the second recording session
APPENDIX B: Corpus tasks and instructions
APPENDIX C: Search for words containing the vocalic sequences of interest
APPENDIX D: Fax samples for the second speaking task
APPENDIX E: Three most relevant parameters per comparison (glottal source analysis)
APPENDIX F: Twin studies in chronological order

Número de págs.: 298
Cómo obtener la tesis: Contactar con la autora
Correo-e del autor/a: <eugeniasansegundogmail.com>

Remitente: Eugenia San Segundo Fernández
Institución: Department of Language and Linguistic Science, University of York
Correo-e: <eugeniasansegundogmail.com>

Fecha: 11 de abril de 2015

Información publicada en Infoling: http://www.infoling.org/informacion/T151.html



Con la ayuda de:
Editorial Arco Libros

© Infoling 1996-2019. Reservados todos los derechos.
ISSN: 1576-3404
Logo image by Hay Kranen / CC-BY