Mervyn A. Jack

Semi-continuous hidden Markov models for speech signals

Abstract A semi-continuous hidden Markov model, which can be considered as a special form of continuous mixture hidden Markov model with the continuous output probability density functions sharing in a mixture Gaussian density codebook, is proposed in this paper. The semicontinuous output probability density function is represented by a combination of the discrete output probabilities of the model and the continuous Gaussian density functions of a mixture Gaussian density codebook. The amount of training data required, as well as the computational complexity of the semi-continuous hidden Markov model, can be significantly reduced in comparison with the continuous mixture hidden Markov model. Parameters of the vector quantization codebook and hidden Markov model can be mutually optimized to achieve an optimal model/codebook combination, which leads to a unified modelling approach to vector quantization and hidden Markov modelling of speech signals. Experimental results are included which show that the recognition accuracy of the semi-continuous hidden Markov model is measurably higher than both the discrete and the continuous hidden Markov model.

User perceptions of security, convenience and usability for ebanking authentication tokens

This research compared three different two-factor methods of eBanking authentication. Three devices employing incremental security layers in the generation of one time passcodes (OTPs) were compared in a repeated-measures, controlled experiment with 50 eBanking customers. Attitudes towards usability and usage logs were taken for each experience. Comparisons of the devices in terms of overall quality, security and convenience as perceived by participants were also recorded. There were significant differences between all three methods in terms of usability measures, perceived quality, convenience and security ratings - with the perceived security ratings following a reverse order to the other measures. Almost two thirds of the participant sample chose the device they perceived the least secure as their preference. Participants were asked to use their preferred method again and tended to find their chosen device more usable. This research illustrates the usability-security trade off, where convenience, quality and usability are sacrificed when increasing layers of security are required. In their preferences, customers were driven by their attitudes towards usability and convenience rather than their perceptions of security.

Scenario-Based Spoken Interaction with Virtual Agents

This paper describes a CALL approach which integrates software for speaker independent continuous speech recognition with embodied virtual agents and virtual worlds to create an immersive environment in which learners can converse in the target language in contextualised scenarios. The result is a self-access learning package: SPELL (Spoken Electronic Language Learning). The SPELL system has been implemented to run in real time on a standard desktop PC. A prototype of SPELL is in user testing with learners of Italian and learners of Japanese in five high schools in Scotland. The speech recogniser is programmed to recognise grammatical and some ungrammatical utterances so that the learner can receive feedback on their language production. The dialogues can be modified in cases of communication difficulties, and are reactive to the learners spoken input so that the learner receives relevant and immediate feedback to their utterance. Feedback takes two key forms: reformulations, where the system modifies the initial input, and recasts, where the system repeats the learners utterance, implicitly correcting any errors. The SPELL system will offer a test bed for the effectiveness of the feedback within a self-access virtual learning environment.

User perceptions of security and usability of single-factor and two-factor authentication in automated telephone banking

This paper describes an experiment to investigate user perceptions of the usability and security of single-factor and two-factor authentication methods in automated telephone banking. In a controlled experiment with 62 banking customers a knowledge-based, single-factor authentication procedure, based on those commonly used in the financial services industry, was compared with a two-factor approach where in addition to the knowledge-based step, a one-time passcode was generated using a hardware security token. Results were gathered on the usability and perceived security of the two methods described, together with call completion rates and call durations for the two methods. Significant differences were found between the two methods, with the two-factor version being perceived as offering higher levels of security than the single-factor authentication version; however, this gain was offset by significantly lower perceptions of usability, and lower ratings for convenience and ease of use for the two-factor version. In addition, the two-factor authentication version took longer for participants to complete. This research provides valuable empirical evidence of the trade-off between security and usability in automated systems.

Usable security: User preferences for authentication methods in eBanking and the effects of experience

Multi-factor authentication involves the use of more than one mode in authentication processes and is typically employed to increase security compared to a fixed password (knowledge-based mode). This research compared three different eBanking authentication processes, a two-layer password (1-factor) method and two alternative 2-factor solutions. The 2-factor processes used One-Time-Passcodes (OTPs) delivered either via a small, single-use device or by text message to a mobile phone. The three authentication methods were compared in a repeated-measures experiment with 141 participants. Three user groups were balanced in the experiment to investigate the effect of experience (current users of the service) on perceptions of usability and security. Attitudes toward usability and observations were taken for each process. Other data gathered quality ratings, preferences and ranked comparisons regarding convenience and security issues. Both 2-factor methods scored significantly higher than the 1-factor method for eBanking authentication usability metrics overall, but experienced users gave higher scores to the 1-factor method they currently use. Overall preferences were spread evenly between the three methods. However, the majority of the participant sample perceived the 1-factor method they had most experience with as being the most secure and most convenient option. The results offer insight into customer attitudes important in their selection of authentication options: convenience, personal ownership and habitual experience of processes.

A usability comparison of three alternative message formats for an SMS banking service

The study reported here compares the usability of three types of message input format: Abbreviations, Numbers and Free-Form as alternatives for a Pull SMS banking service. Participants (N=74) used all three formats to carry out three banking transactions in a repeated measures experiment. The Abbreviations and Numbers versions of the service performed to generally equal levels in all metrics. Free-Form performed the worst as participants took significantly longer to complete tasks and it received significantly lower overall questionnaire and quality scores for satisfaction. The older age group found all three versions in general to be less usable than the younger age group. They took longer on the tasks, had lower completion rates and they also gave each version a lower overall mean satisfaction score.

On the role of metaphor and language in design of third party payments in eBanking: Usability and quality

This paper describes results of a usability study of contrasting user-interface designs for Internet Banking (eBanking). Two specific interface metaphors were compared in the first experiment, linear form filling and array editing interaction modes. Terminology in the interaction dialogue was compared in the second experiment, using typical banking language and a generic, plain language interface. This research aimed to perform usability evaluation and comparison of the alternative interface designs to illuminate the development of new eBanking services. This research involved sixty-one participants (Internet users and customers of the involved Bank) exploring the designs in controlled experiments involving hands-on experience. Banks are interested in ensuring their eBanking services are highly customer-centric and that the interface matches customer expectations in order to drive customers towards this lower cost channel. The results of the first experiment (N=32, where N indicates the number of participants in the cohort) concluded that the simple form-filling metaphor, taken from the traditional paper-based procedure, was generally more usable than a Spreadsheet metaphor. In the second experiment (N=29), it was found that although banking terminology was not completely understood across the cohort, the instructional language changes did not impact significantly on usability.

On the influence of laryngeal pathologies on acoustic and electroglottographic jitter measures

This study compared acoustic and electroglottographic (EGG) jitter from [a] vowels of 103 dysphonic speakers. The EGG recordings were chosen according to their intensity, signal-to-noise ratio, and percentage of unvoiced intervals, while acoustic signals were selected based on voicing detection and the reliability of jitter extraction. The agreement between jitter measures was expressed numerically as a normalized difference. In 63.1% (65/103) of the cases the differences fell within +/-22.5%. Positive differences above +22.5% were associated with increased acoustic jitter and occurred in 12.6% (13/103) of the speakers. These were, typically, cases of small nodular lesions without problems in the posterior larynx. On the other hand, substantial rises in EGG jitter leading to differences below -22.5% took place in 24.3% (25/103) of the speakers and were related to hyperfunctional voices, creaky-like voices, small laryngeal asymmetries affecting the arytenoids, or small-to-moderate glottal chinks. A clinically relevant outcome of the study was the possibility of detecting gentle laryngeal asymmetries among cases of large unilateral increase in EGG jitter. These asymmetries can be linked with vocal problems that are often overlooked in endoscopic examinations.

Experimental assessment of the effectiveness of synthetic personae for multi-modal e-retail applications

This paper details results of an experiment to empirically evaluate the effectiveness and user acceptability of human-like synthetic agents in a multi-modal electronic retail scenario. The synthetic personae played the roles of interactive conversational sales assistants. The range of life-like personae differed with respect to gender and technology. Participants took part in the controlled experiment, which involved them eavesdropping on spoken dialogues between a customer and each of the synthetic personae. They also completed questionnaires and took part in a debriefing interview designed to elicit information relating to the effectiveness, believability and perceived quality of each of the personae. Results show that participants expected a high level of realistic and human-like verbal and non-verbal communicative behaviour in the synthetic personae. This was demonstrated in the strong preference for personae that exhibited natural facial expressions, gestures and emotions. It was also found that disembodied voices were significantly preferred to many of the personae. In addition, results show participants had significantly different attitudes to the voices of the personae.

Unified techniques for vector quantization and hidden Markov modeling using semi-continuous models

A semicontinuous hidden Markov model (HMM), which can be considered as a special form of continuous-mixture HMM with the continuous output probability density functions sharing in a mixture Gaussian density codebook, is proposed. The semicontinuous output probability density function is represented by a combination of the discrete output probabilities of the model and the continuous Gaussian density functions of a mixture Gaussian density codebook. The amount of training data required, as well as the computational complexity of the semicontinuous HMM, can be reduced in comparison to the continuous-mixture HMM. Parameters of the codebook and HMM can be mutually optimized to achieve an optimal model/codebook combination, which leads to a unified modeling approach to vector quantization and hidden Markov modeling of speech signals. Experimental results are included which show that the recognition accuracy of the semicontinuous HMM is measurably higher than those of both the discrete and the continuous HMM.<<ETX>>