Automatic speaker recognition under adverse conditions

Speaker verification is the process of verifying the identity of a person by analysing their speech. There are several important applications for automatic speaker verification (ASV) technology including suspect identification, tracking terrorists and detecting a person’s presence at a remote location in the surveillance domain, as well as person authentication for phone banking and credit card transactions in the private sector. Telephones and telephony networks provide a natural medium for these applications. The aim of this work is to improve the usefulness of ASV technology for practical applications in the presence of adverse conditions. In a telephony environment, background noise, handset mismatch, channel distortions, room acoustics and restrictions on the available testing and training data are common sources of errors for ASV systems. Two research themes were pursued to overcome these adverse conditions: Modelling mismatch and modelling uncertainty. To directly address the performance degradation incurred through mismatched conditions it was proposed to directly model this mismatch. Feature mapping was evaluated for combating handset mismatch and was extended through the use of a blind clustering algorithm to remove the need for accurate handset labels for the training data. Mismatch modelling was then generalised by explicitly modelling the session conditions as a constrained offset of the speaker model means. This session variability modelling approach enabled the modelling of arbitrary sources of mismatch, including handset type, and halved the error rates in many cases. Methods to model the uncertainty in speaker model estimates and verification scores were developed to address the difficulties of limited training and testing data. The Bayes factor was introduced to account for the uncertainty of the speaker model estimates in testing by applying Bayesian theory to the verification criterion, with improved performance in matched conditions. Modelling the uncertainty in the verification score itself met with significant success. Estimating a confidence interval for the "true" verification score enabled an order of magnitude reduction in the average quantity of speech required to make a confident verification decision based on a threshold. The confidence measures developed in this work may also have significant applications for forensic speaker verification tasks.

Repair of calvarial defects with customized tissue-engineered bone grafts - I. Evaluation of osteogenesis in a three-dimensional culture system

Bone generation by autogenous cell transplantation in combination with a biodegradable scaffold is one of the most promising techniques being developed in craniofacial surgery. The objective of this combined in vitro and in vivo study was to evaluate the morphology and osteogenic differentiation of bone marrow derived mesenchymal progenitor cells and calvarial osteoblasts in a two-dimensional (2-D) and three-dimensional (3-D) culture environment (Part I of this study) and their potential in combination with a biodegradable scaffold to reconstruct critical-size calvarial defects in an autologous animal model [Part II of this study; see Schantz, J.T., et al. Tissue Eng. 2003;9(Suppl. 1):S-127-S-139; this issue]. New Zealand White rabbits were used to isolate osteoblasts from calvarial bone chips and bone marrow stromal cells from iliac crest bone marrow aspirates. Multilineage differentiation potential was evaluated in a 2-D culture setting. After amplification, the cells were seeded within a fibrin matrix into a 3-D polycaprolactone (PCL) scaffold system. The constructs were cultured for up to 3 weeks in vitro and assayed for cell attachment and proliferation using phase-contrast light, confocal laser, and scanning electron microscopy and the MTS cell metabolic assay. Osteogenic differentiation was analyzed by determining the expression of alkaline phosphatase (ALP) and osteocalcin. The bone marrow-derived progenitor cells demonstrated the potential to be induced to the osteogenic, adipogenic, and chondrogenic pathways. In a 3-D environment, cell-seeded PCL scaffolds evaluated by confocal laser microscopy revealed continuous cell proliferation and homogeneous cell distribution within the PCL scaffolds. On osteogenic induction mesenchymal progenitor cells (12 U/L) produce significantly higher (p < 0.05) ALP activity than do osteoblasts (2 U/L); however, no significant differences were found in osteocalcin expression. In conclusion, this study showed that the combination of a mechanically stable synthetic framework (PCL scaffolds) and a biomimetic hydrogel (fibrin glue) provides a potential matrix for bone tissue-engineering applications. Comparison of osteogenic differentiation between the two mesenchymal cell sources revealed a similar pattern.