This paper investigates the effect of utterance duration to the calibration of a modern i-vector speaker recognition system with probabilistic linear discriminant analysis (PLDA) modeling. A calibration approach to deal with these effects using quality measure functions (QMFs) is proposed to include duration in the calibration transformation. Extensive experiments are performed in order to evaluate the robustness of the proposed calibration approach for unseen conditions in the training of calibration parameters. Using the latest NIST corpora for evaluation, results highlight the importance of considering the quality metrics like duration in calibrating the scores for automatic speaker recognition systems.

No abstract

Motivated by the application of speaker recognition in forensic area, this paper presents a study on noise robustness of several automatic speaker recognition system approaches, ranging from simple dotscoring and a standard i-vector system with cosine distance scoring to a state-of-the-art i-vector Probabilistic Linear Discriminant Analysis (PLDA) system. Using the recent NIST 2010 Speaker Recognition Evaluation (SRE) data, the systems are analyzed in added noise conditions with a range of signal to noise ratios. Various experiments were conducted to study the influence of the noise on the speech activity detection and Wiener filtering in the front-end of the system.

An evaluation of the verification and calibration performance of a face recognition system based on inter-session variability modelling is presented. As an extension to calibration through linear transformation of scores, categorical calibration is introduced as a way to include additional information about images for calibration. The cost of likelihood ratio, which is a well-known measure in the speaker recognition field, is used as a calibration performance metric. The results obtained from the challenging mobile biometrics and surveillance camera face databases indicate that linearly calibrated face recognition scores are less misleading in their likelihood ratio interpretation than uncalibrated scores. In addition, the categorical calibration experiments show that calibration can be used not only to improve the likelihood ratio interpretation of scores, but also to improve the verification performance of a face recognition system.

Post-stroke patients need ongoing rehabilitation to restore dysfunction caused by an attack so that a monitoring device is required. EEG signals reflect electrical activity in the brain, which also informs the condition of post-stroke patient recovery. However, the EEG signal processing model needs to provide information on the post-stroke state. The development of deep learning allows it to be applied to the identification of post-stroke patients. This study proposed a method for identifying post-stroke patients using convolutional neural networks (CNN). Wavelet is used for EEG signal information extraction as a feature of machine learning, which reflects the condition of post-stroke patients. This feature is Delta, Alpha, Beta, Theta, and Mu waves. Moreover, the five waves, amplitude features are also added according to the characteristics of the post-stroke EEG signal. The results showed that the feature configuration is essential as distinguish. The accuracy of the testing data was 90% with amplitude and Beta features compared to 70% without amplitude or Beta. The experimental results also showed that adaptive moment estimation (Adam) optimization model was more stable compared to Stochastic gradient descent (SGD). But SGD can provide higher accuracy than the Adam model.