Finding meta data in speech and handwriting biometrics

The goal of this paper is to present our work on the analysis of speech and handwriting biometrics related to meta data, which are based on one side on system hardware specifics (technical meta data) and on the other side to personal attributes (non-technical meta data). System related meta data represent physical characteristics of biometric sensors and are essential for ensuring comparable quality of the biometric raw signals. Previous work in personal related meta data has shown that it is possible to estimate some meta data like script language, dialect, origin, gender and age by statistically analyzing human handwriting and voice data. On one hand, by knowing both kinds of such meta data, it appears to be possible to adapt the recognition or authentication algorithms in order to enhance their recognition accuracy and to analyze the sensor dependency of biometric algorithms with respect to hardware properties such as sampling resolution. On the other hand, interesting aspects are to evaluate, if cultural characteristics (such as native language, or ethnicity) can be derived by statistical or analytical means from voice or handwriting dynamics and to which degree grouping of users by persons with identical or similar meta data may result in better biometric recognition accuracy. All these aspects have been widely neglected by research until today. This article will discuss approaches to model such meta data and strategies for finding features by introducing a new meta data taxonomy, from which we derive those personal and system attributes related to the cultural background, which are employed in our experimental evaluation. Further, we describe the test methodology used for our experimental evaluation in different cultural regions of India and Europe and present first results for sensor hardware related meta data in handwriting biometrics as well as language related meta data in speaker recognition.