Extending the task of diarization to speaker attribution

In this paper we extend the concept of speaker annotation within a single-recording, or speaker diarization, to a collection wide approach we call speaker attribution. Accordingly, speaker attribution is the task of clustering expectantly homogenous intersession clusters obtained using diarization according to common cross-recording identities. The result of attribution is a collection of spoken audio across multiple recordings attributed to speaker identities. In this paper, an attribution system is proposed using mean-only MAP adaptation of a combined-gender UBM to model clusters from a perfect diarization system, as well as a JFA-based system with session variability compensation. The normalized cross-likelihood ratio is calculated for each pair of clusters to construct an attribution matrix and the complete linkage algorithm is employed to conduct clustering of the inter-session clusters. A matched cluster purity and coverage of 87.1% was obtained on the NIST 2008 SRE corpus.

Partial protection against chlamydial reproductive tract infection by a recombinant major outer membrane protein/CpG/cholera toxin intranasal vaccine in the guinea pig Chlamydia cavaie model

Chlamydia trachomatis is a major cause of sexually transmitted diseases worldwide. There currently is no vaccine to protect against chlamydial infection of the female reproductive tract. Vaccine development has predominantly involved using the murine model, however infection of female guinea pigs with Chlamydia caviae more closely resembles chlamydial infection of the human female reproductive tract, and presents a better model to assess potential human chlamydial vaccines. We immunised female guinea pigs intranasally with recombinant major outer membrane protein (r-MOMP) combined with CpG-10109 and cholera toxin adjuvants. Both systemic and mucosal immune responses were elicited in immunised animals. MOMP-specific IgG and IgA were present in the vaginal mucosae, and high levels of MOMP-specific IgG were detected in the serum of immunised animals. Antibodies from the vaginal mucosae were also shown to be capable of neutralising C. caviae in vitro. Following immunisation, animals were challenged intravaginally with a live C. caviae infection of 102 inclusion forming units. We observed a decrease in duration of infection and a significant (p<0.025) reduction in infection load in r-MOMP immunised animals, compared to animals immunised with adjuvant only. Importantly, we also observed a marked reduction in upper reproductive tract (URT) pathology in r-MOMP immunised animals. Intranasal immunisation of female guinea pigs with r-MOMP was able to provide partial protection against C. caviae infection, not only by reducing chlamydial burden but also URT pathology. This data demonstrates the value of using the guinea pig model to evaluate potential chlamydial vaccines for protection against infection and disease pathology caused by C. trachomatis in the female reproductive tract.

A comparison of the effects of a chlamydial vaccine administered during or after a C. muridarum urogenital infection of female mice

There are approximately 92 million new chlamydial infections of the genital tract in humans diagnosed each year, costing health care systems billions of dollars in treatment not only of acute infections, but also of associated inflammatory sequelae, such as pelvic inflammatory disease (PID) and ectopic pregnancy. These numbers are increasing at a steady rate and, due to the asymptomatic nature of infections, the incidence may be underestimated and the costs of treatment therefore higher. Over the previous few decades there has been a large amount of research into the development of an efficacious vaccine against genital tract chlamydial infections. The majority of this research has focused on females, due to the high rate of development of associated diseases, including PID, which can lead to ectopic pregnancy and infertility. In light of the increasing infection rates that have occurred despite the availability of antibiotics, and the asymptomatic nature of chlamydial infections, it is imperative that an efficacious vaccine that protects against infection and associated pathology be developed.

Can audio-visual speech recognition outperform acoustically enhanced speech recognition in automotive environment?

The use of visual features in the form of lip movements to improve the performance of acoustic speech recognition has been shown to work well, particularly in noisy acoustic conditions. However, whether this technique can outperform speech recognition incorporating well-known acoustic enhancement techniques, such as spectral subtraction, or multi-channel beamforming is not known. This is an important question to be answered especially in an automotive environment, for the design of an efficient human-vehicle computer interface. We perform a variety of speech recognition experiments on a challenging automotive speech dataset and results show that synchronous HMM-based audio-visual fusion can outperform traditional single as well as multi-channel acoustic speech enhancement techniques. We also show that further improvement in recognition performance can be obtained by fusing speech-enhanced audio with the visual modality, demonstrating the complementary nature of the two robust speech recognition approaches.

Cross likelihood ratio based speaker clustering using eigenvoice models

This paper proposes the use of eigenvoice modeling techniques with the Cross Likelihood Ratio (CLR) as a criterion for speaker clustering within a speaker diarization system. The CLR has previously been shown to be a robust decision criterion for speaker clustering using Gaussian Mixture Models. Recently, eigenvoice modeling techniques have become increasingly popular, due to its ability to adequately represent a speaker based on sparse training data, as well as an improved capture of differences in speaker characteristics. This paper hence proposes that it would be beneficial to capitalize on the advantages of eigenvoice modeling in a CLR framework. Results obtained on the 2002 Rich Transcription (RT-02) Evaluation dataset show an improved clustering performance, resulting in a 35.1% relative improvement in the overall Diarization Error Rate (DER) compared to the baseline system.

i-vector based speaker recognition on short utterances

Robust speaker verification on short utterances remains a key consideration when deploying automatic speaker recognition, as many real world applications often have access to only limited duration speech data. This paper explores how the recent technologies focused around total variability modeling behave when training and testing utterance lengths are reduced. Results are presented which provide a comparison of Joint Factor Analysis (JFA) and i-vector based systems including various compensation techniques; Within-Class Covariance Normalization (WCCN), LDA, Scatter Difference Nuisance Attribute Projection (SDNAP) and Gaussian Probabilistic Linear Discriminant Analysis (GPLDA). Speaker verification performance for utterances with as little as 2 sec of data taken from the NIST Speaker Recognition Evaluations are presented to provide a clearer picture of the current performance characteristics of these techniques in short utterance conditions.

Marginalised students and insider knowledge

One of the claims made for valuing the voices of marginalised students is that an insider perspective can be revealed on student issues and the ways in which education policies and systems impact on them. This chapter examines the ways in which participants in an Australian ‘students-as-researchers’ (SaR) project were able to raise knowledge of and address, to some extent, long-standing issues of racism in their schools. The SaR project has operated in more than thirty schools for periods of one to five years. Based on a participatory action research model, groups of secondary school students from schools serving socio-economically disadvantaged communities have worked with nominated teachers and university researchers to identify and research local issues relating to low academic outcomes and to develop and enact responses to the identified concerns. The voices of marginalised students quoted in this chapter illustrate that important insider knowledge can be revealed through the SaR process. Where student views have been acknowledged and acted on by the schools, significant change to student-teacher relationships and school culture has been achieved; the participants have been personally empowered and academic improvements across the schools have been noted. For such change to occur, however, a culture of mutual respect must be created in which teachers and school administrators value students’ views and are open to the possibility of unfavourable criticism.

Visual voice activity detection using frontal versus profile views

Visual activity detection of lip movements can be used to overcome the poor performance of voice activity detection based solely in the audio domain, particularly in noisy acoustic conditions. However, most of the research conducted in visual voice activity detection (VVAD) has neglected addressing variabilities in the visual domain such as viewpoint variation. In this paper we investigate the effectiveness of the visual information from the speaker’s frontal and profile views (i.e left and right side views) for the task of VVAD. As far as we are aware, our work constitutes the first real attempt to study this problem. We describe our visual front end approach and the Gaussian mixture model (GMM) based VVAD framework, and report the experimental results using the freely available CUAVE database. The experimental results show that VVAD is indeed possible from profile views and we give a quantitative comparison of VVAD based on frontal and profile views The results presented are useful in the development of multi-modal Human Machine Interaction (HMI) using a single camera, where the speaker’s face may not always be frontal.

Investigative thinking and creativity : an empirical study of police detectives in Norway

This paper makes distinctions among four investigative thinking styles of detectives: method style, challenge style, skill style, and risk style. Based on previous research, this study empirically tested, to what extent there are cumulative relationships among these thinking styles. Furthermore, this research studied relationships between investigative thinking styles and creativity in police investigations. Significant relationships were found between the extent of the challenge and risk styles and the extent of creativity.

Knowledge management in law enforcement : knowledge views for patrolling police officers

This paper describes and classifies different types of knowledge that are a part of police patrol officer's practice. Even though an investigation usually forces a police officer to apply several different knowledge types, this paper discusses different forms of professional knowledge separately to enable categorisation.