Evidentiary Basis Development with Support of Forensic Research Assessment

Increased attention to the development of modern approaches to probabilistic evaluation of research results in the field of forensic science is attributed to the trends of critical analysis of the current state of forensic science and requirements for clear characteristics of limitations of research results, including indicators of uncertainty of the data obtained and associated estimated probabilities. One of the fundamental provisions of the modern theory of evaluating judicial evidence, along with the admissibility and desirability of using probabilities, is the principle of comparing these probabilities in the light of their conditionality by competing versions arising from the adversarial nature of justice. In this regard, the purpose of this article is to develop methodological approaches to the use of the likelihood ratio as the most appropriate form of determining the significance of conclusions sent by an expert to the court for the formation of evidence. The empirical basis of the article is based on a brief review of publications from 2000 to 2018, devoted to the application of the concept of the likelihood ratio in forensic activities. According to many scientists, the use of this concept can provide a real assessment of the reliability of the evidence. In legal proceedings, evidence is generally understood as information about facts obtained in accordance with the procedure provided for by law, on the basis of which the presence or absence of circumstances that are important for the proper consideration and resolution of the case is established. In this publication, the term «evidence» is considered through an expert-technological prism and is presented as various quantitative continuous measurements (properties and characteristics of objects of forensic expertise), which are used when comparing a known and questioned sample to solve the question of their origin from one or from different sources. The article discusses the most common normal distribution of continuous data and a general approach to calculating the likelihood ratio (LR) using probability density functions (pdf). It is shown that in order to account for the variability of compared samples, three databases are required for calculating LR: a potential database, a control database of a known sample, and a comparative database of a questioned sample. Examples of calculating the LR and strength of evidence performed for various types of examinations are given. The procedures for calculating LR are generally the same, but the authors suggest different techniques to calculate and graphically represent the strength of the proof. In more detail, the publications present the so-called value of the cost or penalty for an incorrect forecast (C_lLR), introduced the terms of trueness and reproducibility, as well as the confidence interval of this value. The article highlights a number of features of calculating LR for multidimensional continuous data. Of great interest is the use of the speaker model in sound recording expertise in the form of a weighted sum of Gaussian densities M components (Gaussian mixture models — GMM). Each density component in this sum is a D-dimensional Gaussian pdf with an average vector value and a covariance matrix. It can be assumed that the use of GMM-pdf in LR calculations is effective not only for forensic examination of speaker recognition, but also for other types of examinations. The universality of assessing the similarity/difference of objects of forensic research using the likelihood ratio indicates the prospects for applying the concept.

1. Evett J. W. Towards a uniform framework for reporting opinions in forensic science case-work / Sci. Justice. 1998. Vol. 38. N 3. P. 198 – 202. DOI: 10.1016/S1355-0306(98)72105-7.

2. Champod C., Meuwly D. The inference of identity in forensic speaker recognition / Speech Comm. 2000. Vol. 31. P. 193 – 203. DOI: 10.1016/S0167-6393(99)00078-3.

3. Koons R. D., Buscaglia J. Interpretation of glass composition measurements: the effects of match criteria on discrimination capability / J. Forensic Sci. 2001. Vol. 47. N 3. P. 505 – 512. DOI: 10.1520/JFS2001349.

4. Aitken C. G. G., Lucy D. Evaluation of trace evidence in the form of multivariate data / Appl. Statist. 2004. Vol. 53. Part 1. P. 109 – 122. DOI: 10.26896/1028-6861-2018-84-6-70-76.

5. Aitken C. G. G., Taroni F. Statistics and the Evaluation of Forensic Evidence for Forensic Scientist. — London: Wiley, 2004. — 510 p.

6. Botti F., Alexander A., Drygajlo A. On compensation of mismatched recording conditions in the Bayesian approach for forensic automatic speaker recognition / Forensic Sci. Int. 2004. Vol. 146. Suppl. 2. P. S101 – S106. DOI: 10.1016/j.forsciint.2004.09.032.

7. Balding D. J. Weight-of-evidence for Forensic DNA Profiles. — London: Wiley, 2005. — 198 p.

8. Curran M. An introduction to Bayesian credible intervals for sampling error in DNA profiles / Law, Probability and Risk. 2005. Vol. 4. P. 115 – 126. DOI: 10.1093/lpr/mgi009.

9. González-Rodríguez J., Drygajlo A. V., Ramos-Castro D., et al. Robust estimation, interpretation and assessment of likelihood ratios in forensic speaker recognition / Comput. Speech Lang. 2006. Vol. 20. P. 331 – 355. DOI: 10.1016/j.csl.2005.08.005.

10. Rose P. Technical forensic speaker recognition: evaluation, types and testing of evidence / Comput. Speech Lang. 2006. Vol. 20. N 2 – 3. P. 159 – 191. DOI: 10.1016/j.csl.2005.07.003.

11. Morrison G. S. Forensic voice comparison and the paradigm shift / Sci. Justice. 2009. Vol. 49. P. 298 – 308. DOI: 10.1016/j.scijus.2009.09.002.

12. Association of Forensic Science Providers. Standards for the formulation of evaluative forensic science expert opinion / Sci. Justice. 2009. Vol. 49. P. 161 – 164. DOI: 10.1016/j.scijus.2009.07.004.

13. Rose P., Morrison G. S. A response to the UK position statement on forensic speaker comparison / Int. J. Speech Lang. Law 2009. Vol. 16. P. 139 – 163. DOI: 10.1558/ijsll.v16i1.139.

14. Taroni F., Bozza S., Biedermann A., et al. Data Analysis in Forensic Science: A Bayesian Decision Perspective. — NY: Wiley, 2010. — 390 p.

15. Zadora G., Ramos D. Evaluation of glass samples for forensic purposes — an application of likelihood ratio model and information-theoretical approach / Chemom. Intell. Lab. Syst. 2010. Vol. 102. P. 63 – 68. DOI: 10.1016/j.chemolab.2010.007.

16. Morrison G. Measuring the validity and reliability of forensic likelihood-ratio systems / Sci. Justice. 2011. Vol. 51. N 3. P. 91 – 98. DOI: 10.1016/j.scijus.2011.03.002.

17. Neumann C., Evett J. W., Skerrett J. Quantifying the weight of evidence from a forensic fingerprint comparison: a new paradigm / J. R. Stat. Soc. Ser. A. 2012. Vol. 175. N 2. P. 371 – 415. DOI: 10.1111/j.1467-985X.2011.01027.x.

18. Bebeshko G. I., Omelyuk G. G., Usov A. I. The role and significance of likelihood ratio concept in assessment and interpretation of the results of forensic activities / Zavod. Lab. Diagn. Mater. 2018. Vol. 84. N 6. P. 70 – 76. DOI: 10.26896/1028-6861-2018-84-6-70-76 [in Russian].

19. Gradusova O. V., Kuz’min S. A. Probability Interpretation of Forensic Evidence / Teor. Prakt. Sud. Ékspert. 2017. Vol. 12. N 4. P. 27 – 40 [in Russian].

20. Nefedov S. N. Bayesian approach to evidence assessment and standardization of verbal statements of expert conclusions. N 8. — Minsk: RIPO, 2015. P. 187 – 195 [in Russian].

21. Lindley D. V. A problem in forensic science / Biometrika. 1977. Vol. 64. P. 207 – 213.

22. Curran J. M., Triggs C. M., Almirall J. R., et al. The interpretation of elemental composition from forensic glass evidence: (I and II) / Sci. Justice. 1997. Vol. 37. N 4. P. 241 – 244, 245 – 249. DOI: 10.1016/S1355-0305(97)72197-X, 10.1016/S1355-0306(97)72198-1.

23. Koons R. D., Buscaglia J. Interpretation of glass composition measurements: the effects of match criteria on discrimination capability / J. Forensic Sci. 2002. Vol. 47. P. 505 – 512.

24. Martyna A., Lucy D., Zadora G. V., et al. The evidential value of microspectrophotometry measurements made for pen inks / Anal. Methods. 2013. Vol. 5. P. 6788 – 6795. DOI: 10.1039/C341622D.

25. Brümmer N., du Preez J. Application independent evaluation of speaker detection / Comp. Speech Lang. 2006. Vol. 20. N 2 – 3. P. 230 – 275. DOI: 10.1016/j.csl.2005.08.001.

26. Ramos D., Gonzalez-Rodriguez J. Reliable support: Measuring calibration of likelihood ratios / Forensic Sci. Int. 2013. Vol. 230. N 1 – 3. P. 156 – 169. DOI: 10.1016/j.forsciint.2013.04.014.

27. Ramos D., Gonzalez-Rodriguez J., Zadora G., Aitken C. Information-theoretical assessment of the performance of likelihood ratio computation methods / J. Forensic Sci. 2013. Vol. 58. N 6. P. 1503 – 1518. DOI: 10.1111/1556-4029.12233.

28. Drygajlo A. Automatic Speaker Recognition for Forensic Case Assessment and Interpretation. — In: A. Neustein, H. A. Patil, Eds. Forensic Speaker Recognition. Ch. 2. 2012. P. 21 – 39. DOI: 10.1007/978-1-4614-0263-3_2.

29. Kinnunen T., Li H. An overview of text-independent speaker recognition: from features to supervectors / Speech Commun. 2010. Vol. 52. N 1. P. 12 – 40. DOI: 10.1016/j.specom.2009.08.009.

30. Meuwly D., Drygajlo A. Forensic Speaker Recognition Based on a Bayesian Framework and Gaussian Mixture Modelling (GMM) / The Speaker Recognition Workshop. Crete, Greece, June 18 – 22. A Speaker Odyssey. 2001. — 6 p.

31. Reynolds D. A. Automatic Speaker Recognition Using Gaussian Mixture Speaker Models / The Lincoln Lab. J. 1995. Vol. 8. N 2. P. 173 – 191.

32. Reynolds D. A., Quatieri T. F., Dunn R. B. Speaker Verification Using Adapted Gaussian Mixture Models / Digital Signal Proc. 2000. Vol. 10. N. 1 – 3. P. 19 – 41. DOI: 10.1006/dspr.1999.0361.

33. Matveev Yu. N. Technologies for biometric identification of individuals by voice and other modalities / Vestnik MGTU im. N. Й. Baumana. 2012. P. 46 – 61 [in Russian].

34. Kozlov A. V., Kudashev O. Yu., Matveev, et al. Speaker recognition system for the NISTSRE / SPIIRAS Proc. 2013. N 2(25). P. 350 – 370 [in Russian].

35. Morrison G. S., Zhang C., Rose P. An empirical estimate of the precision of likelihood ratios from a forensic-voice-comparison system / Forensic Sci. Int. 2011. Vol. 208. P. 59 – 65. DOI: 10.1016/j.forsciint.2010.11.001.

36. González-Rodríguez J., Rose P., Ramos D., et al. Emulating DNA: rigorous quantification of evidential weight in transparent and testable forensic speaker recognition / IEEE Trans. Audio Speech Lang. Process. 2007. Vol. 15. P. 2104 – 2115. DOI: 10.1109/TASL.2007.902747.

37. Morrison G. S. Likelihood-ratio forensic voice comparison using parametric representations of the formant trajectories of diphthongs / J. Acoust. Soc. Am. 2009. Vol. 125. P. 2387 – 2397. DOI: 10.1121/1.3081384.

38. Van Leeuwen D. A., Brümmer N. An introduction to application-independent evaluation of speaker recognition systems. — In the book: C. Müller (ed.), Speaker Classification I: Fundamentals, Features, and Methods. — Heidelberg, Germany: Springer-Verlag, 2007. P. 330 – 353. DOI: 10.1007/978-3-540-74200-5_19.

39. Morrison G. S. Forensic voice comparison using likelihood ratios based on polynomial curves fitted to the formant trajectories of Australian English (aI) / Int. J. Speech Lang. Law. 2008. Vol. 15. P. 247 – 264. DOI: 10.1558/ijsll.v15i2.249.

40. Morrison G. S. A comparison of procedures for the calculation of forensic likelihood ratios from acoustic-phonetic data: multivariate kernel density (MVKD) versus Gaussian mixture model — universal background model (GMM-UBM) / Speech Commun. 2011. Vol. 53. P. 242 – 256. DOI: 10.1016/j.specom.2010.09.005.

41. Aitken C. G. G. Statistical discriminant analysis in forensic science / J. Forensic Sci. Soc. 1986. Vol. 26. P. 237 – 247.

42. Berry D. A., Evett I. W., Pinchin R. Statistical inference in crime investigations using deoxyribonucleic acid profiling (with discussion) / Appl. Statist. 1992. Vol. 41. P. 499 – 531. DOI: 10.1111/j.1467-9876.1992/tb02418.x.

43. Chan K. P. S., Aitken C. G. G. Estimation of the Bayes’ factor in a forensic science problem / J. Statist. Computn. Simuln. 1989. Vol. 33. P. 249 – 264.

44. Brümmer N., Burget L., Cernocký J. H., et al. Fusion of heterogenous speaker recognition systems in the STBU submission for the NIST SRE 2006 / IEEE Trans. Audio Speech Lang. Process. 2007. Vol. 15. P. 2072 – 2084. DOI: 10.1109/TASL.2007.902870.

45. Pigeon S., Druyts P., Verlinde P. Applying logistic regression to the fusion of the NIST’99 1-speaker submissions / Digital Signal Proc. 2000. Vol. 10. P. 237 – 248. DOI: 10.1006/dspr.1999.0358.

46. Martyna A., Michalska A., Zadora G. Interpretation of FTIR spectra of polymers and Raman spectra of car paints by means of likelihood ratio approach supported by wavelet transform for reducing data dimensionality / Anal. Bioanal. Chem. 2015. Vol. 407. P. 3357 – 3376. DOI: 10.1007/s00216-015-8558-9.

47. Daubechies I. Ten Lectures on Wavelets — Philadelphia: CBMS-NSF. Regional Conference Series in Applied Mathematics, 1992.