The effect of dialect mismatch on likelihood-maximising speech enhancement for noise-robust speech recognition

Traditional speech enhancement methods optimise signal-level criteria such as signal-to-noise ratio, but these approaches are sub-optimal for noise-robust speech recognition. Likelihood-maximising (LIMA) frameworks are an alternative that optimise parameters of enhancement algorithms based on state sequences generated for utterances with known transcriptions. Previous reports of LIMA frameworks have shown significant promise for improving speech recognition accuracies under additive background noise for a range of speech enhancement techniques. In this paper we discuss the drawbacks of the LIMA approach when multiple layers of acoustic mismatch are present – namely background noise and speaker accent. Experimentation using LIMA-based Mel-filterbank noise subtraction on American and Australian English in-car speech databases supports this discussion, demonstrating that inferior speech recognition performance occurs when a second layer of mismatch is seen during evaluation.

The application of phonetic distribution normalisation to likelihood-maximising speech enhancement for robust ASR

Traditional speech enhancement methods optimise signal-level criteria such as signal-to-noise ratio, but such approaches are sub-optimal for noise-robust speech recognition. Likelihood-maximising (LIMA) frameworks on the other hand, optimise the parameters of speech enhancement algorithms based on state sequences generated by a speech recogniser for utterances of known transcriptions. Previous applications of LIMA frameworks have generated a set of global enhancement parameters for all model states without taking in account the distribution of model occurrence, making optimisation susceptible to favouring frequently occurring models, in particular silence. In this paper, we demonstrate the existence of highly disproportionate phonetic distributions on two corpora with distinct speech tasks, and propose to normalise the influence of each phone based on a priori occurrence probabilities. Likelihood analysis and speech recognition experiments verify this approach for improving ASR performance in noisy environments.

Self-service technology and the moderating effects of consumer characteristics

The incidence of self-service technology, where the consumer delivers the service themselves using technology, is increasing in the service encounter. One area that is under-explored is the potential impact of self-service technology on consumer satisfaction and affective commitment. Accordingly, this paper presents an empirical study that investigates the relative impact of self-service technology on consumer satisfaction (both overall and transaction-specific) and affective commitment, accounting for the moderating effects of consumer characteristics. The results highlight the importance of personal service for evaluations of satisfaction and commitment, and the importance of social competency as a moderator in this relationship. An understanding of these consumer perceptions will allow organisations to develop strategies to deliver the services expected by their consumers, improving consumer satisfaction and commitment.

Recognising audio-visual speech in vehicles using the AVICAR database

Interacting with technology within a vehicle environment using a voice interface can greatly reduce the effects of driver distraction. Most current approaches to this problem only utilise the audio signal, making them susceptible to acoustic noise. An obvious approach to circumvent this is to use the visual modality in addition. However, capturing, storing and distributing audio-visual data in a vehicle environment is very costly and difficult. One current dataset available for such research is the AVICAR [1] database. Unfortunately this database is largely unusable due to timing mismatch between the two streams and in addition, no protocol is available. We have overcome this problem by re-synchronising the streams on the phone-number portion of the dataset and established a protocol for further research. This paper presents the first audio-visual results on this dataset for speaker-independent speech recognition. We hope this will serve as a catalyst for future research in this area.

JFA based Speaker Recognition using Delta-Phase and MFCC features

This paper investigates the use of mel-frequency deltaphase (MFDP) features in comparison to, and in fusion with, traditional mel-frequency cepstral coefficient (MFCC) features within joint factor analysis (JFA) speaker verification. MFCC features, commonly used in speaker recognition systems, are derived purely from the magnitude spectrum, with the phase spectrum completely discarded. In this paper, we investigate if features derived from the phase spectrum can provide additional speaker discriminant information to the traditional MFCC approach in a JFA based speaker verification system. Results are presented which provide a comparison of MFCC-only, MFDPonly and score fusion of the two approaches within a JFA speaker verification approach. Based upon the results presented using the NIST 2008 Speaker Recognition Evaluation (SRE) dataset, we believe that, while MFDP features alone cannot compete with MFCC features, MFDP can provide complementary information that result in improved speaker verification performance when both approaches are combined in score fusion, particularly in the case of shorter utterances.

Speaker linking using complete-linkage clustering

Speaker diarization determines instances of the same speaker within a recording. Extending this task to a collection of recordings for linking together segments spoken by a unique speaker requires speaker linking. In this paper we propose a speaker linking system using linkage clustering and state-of-the-art speaker recognition techniques. We evaluate our approach against two baseline linking systems using agglomerative cluster merging (AC) and agglomerative clustering with model retraining (ACR). We demonstrate that our linking method, using complete-linkage clustering, provides a relative improvement of 20% and 29% in attribution error rate (AER), over the AC and ACR systems, respectively.

Past 140-year environmental record in the northern South China Sea : evidence from coral skeletal trace metal variations

About 140-year changes in the trace metals in Porites coral samples from two locations in the northern South China Sea were investigated. Results of PCA analyses suggest that near the coast, terrestrial input impacted behavior of trace metals by 28.4%, impact of Sea Surface Temperature (SST) was 19.0%, contribution of war and infrastructure were 14.4% and 15.6% respectively. But for a location in the open sea, contribution of War and SST reached 33.2% and 16.5%, while activities of infrastructure and guano exploration reached 13.2% and 14.7%. While the spatiotemporal change model of Cu, Cd and Pb in seawater of the north area of South China Sea during 1986–1997 were reconstructed. It was found that in the sea area Cu and Cd contaminations were distributed near the coast while areas around Sanya, Hainan had high Pb levels because of the well-developed tourism related activities.

Impinging jet simulation of stationary downburst flow over topography

A non-translating, long duration thunderstorm downburst has been simulated experimentally and numerically by modelling a spatially stationary steady flow impinging air jet. Velocity profiles were shown to compare well with an upper-bound of velocity measurements reported for full-scale microbursts. Velocity speed-up over a range of topographic features in simulated downburst flow was also tested with comparisons made to previous work in a similar flow, and also boundary layer wind tunnel experiments. It was found that the amplification measured above the crest of topographic features in simulated downburst flow was up to 35% less than that observed in boundary layer flow for all shapes tested. From the computational standpoint we conclude that the Shear Stress Transport (SST) model performs the best from amongst a range of eddy-viscosity and second moment closures tested for modelling the impinging jet flow.

Discerning the timing and cause of historical mortality events in modern Porites from the Great Barrier Reef

The life history strategies of massive Porites corals make them a valuable resource not only as key providers of reef structure, but also as recorders of past environmental change. Yet recent documented evidence of an unprecedented increase in the frequency of mortality in Porites warrants investigation into the history of mortality and associated drivers. To achieve this, both an accurate chronology and an understanding of the life history strategies of Porites are necessary. Sixty-two individual Uranium–Thorium (U–Th) dates from 50 dead massive Porites colonies from the central inshore region of the Great Barrier Reef (GBR) revealed the timing of mortality to have occurred predominantly over two main periods from 1989.2 ± 4.1 to 2001.4 ± 4.1, and from 2006.4 ± 1.8 to 2008.4 ± 2.2 A.D., with a small number of colonies dating earlier. Overall, the peak ages of mortality are significantly correlated with maximum sea-surface temperature anomalies. Despite potential sampling bias, the frequency of mortality increased dramatically post-1980. These observations are similar to the results reported for the Southern South China Sea. High resolution measurements of Sr/Ca and Mg/Ca obtained from a well preserved sample that died in 1994.6 ± 2.3 revealed that the time of death occurred at the peak of sea surface temperatures (SST) during the austral summer. In contrast, Sr/Ca and Mg/Ca analysis in two colonies dated to 2006.9 ± 3.0 and 2008.3 ± 2.0, suggest that both died after the austral winter. An increase in Sr/Ca ratios and the presence of low Mg-calcite cements (as determined by SEM and elemental ratio analysis) in one of the colonies was attributed to stressful conditions that may have persisted for some time prior to mortality. For both colonies, however, the timing of mortality coincides with the 4th and 6th largest flood events reported for the Burdekin River in the past 60 years, implying that factors associated with terrestrial runoff may have been responsible for mortality. Our results show that a combination of U–Th and elemental ratio geochemistry can potentially be used to precisely and accurately determine the timing and season of mortality in modern massive Porites corals. For reefs where long-term monitoring data are absent, the ability to reconstruct historical events in coral communities may prove useful to reef managers by providing some baseline knowledge on disturbance history and associated drivers.

Short utterance PLDA speaker verification using SN-WLDA and variance modelling techniques

This paper proposes a combination of source-normalized weighted linear discriminant analysis (SN-WLDA) and short utterance variance (SUV) PLDA modelling to improve the short utterance PLDA speaker verification. As short-length utterance i-vectors vary with the speaker, session variations and phonetic content of the utterance (utterance variation), a combined approach of SN-WLDA projection and SUV PLDA modelling is used to compensate the session and utterance variations. Experimental studies have found that a combination of SN-WLDA and SUV PLDA modelling approach shows an improvement over baseline system (WCCN[LDA]-projected Gaussian PLDA (GPLDA)) as this approach effectively compensates the session and utterance variations.

Theorizing Masculinity in Information Systems Research

INTRODUCTION: I want to argue that understanding masculinity is an important part of understanding gender and sexuality as it relates to information and communications technologies (ICTs), specifically those under the lens of the information-systems community. In order to do this, the landscape of gender and sexuality research in general is referred to along with such research in the field of information systems (IS), with reference as necessary to masculinity studies. I will then suggest some possible areas where a more thorough going theorization may prove useful. In sum, future research might focus on the relation-ship between marginalised masculinities and the construction and consumption of is in work organisations and society…

Phonetic spoken term search using topic information

The aim of spoken term detection (STD) is to find all occurrences of a specified query term in a large audio database. This process is usually divided into two steps: indexing and search. In a previous study, it was shown that knowing the topic of an audio document would help to improve the accuracy of indexing step which results in a better performance for STD system. In this paper, we propose the use of topic information not only in the indexing step, but also in the search step. Results of our experiments show that topic information could also be used in search step to improve the STD accuracy.

Rescaling clustering trees using impact ratios for robust hierarchical speaker clustering

We present a novel method for improving hierarchical speaker clustering in the tasks of speaker diarization and speaker linking. In hierarchical clustering, a tree can be formed that demonstrates various levels of clustering. We propose a ratio that expresses the impact of each cluster on the formation of this tree and use this to rescale cluster scores. This provides score normalisation based on the impact of each cluster. We use a state-of-the-art speaker diarization and linking system across the SAIVT-BNEWS corpus to show that our proposed impact ratio can provide a relative improvement of 16% in diarization error rate (DER).

Consumer’s persistence in solving their own problem with self‐service technology

An increasing range of technology services are now offered on a self-service basis. However, problems with self-service technologies (SSTs) occur at times due to the technical error, staff error, or consumers’ own mistakes. Considering the role of consumers as co-producers in the SST context, we aim to study consumer’s behaviours, strategies, and decision making in solving their problem with SST and identify the factors contributing to their persistence in solving the problem. This study contributes to the information systems research, as it is the first study that aims to identify such a process and the factors affecting consumers’ persistence in solving their problem with SST. A focus group with user support staff has been conducted, yielding some initial results that helped to conduct the next phases of the study. Next, using Critical Incident Technique, data will be gathered through focus groups with users, diary method, and think-aloud method.

Investigating the factors influencing the adoption of m-banking: A cross cultural study

Purpose Little is known about the adoption of mobile banking technologies in emerging Asian economies. This paper aims to empirically examine the motivators that influence a consumer’s intentions to use mobile banking. Design/methodology/approach A web-based survey was employed to collect data from 348 respondents, split across Thailand and Australia. Data were analyzed by employing exploratory and confirmatory factor analyses, path and invariance analyses. Findings The findings indicate that for Australian consumers, perceived ease of use, perceived usefulness and perceived risk were the primary determinants of mobile banking adoption. For Thai consumers, the main factors were perceived usefulness, perceived risk and social influence. National culture was found to impact key antecedents that lead to adoption of m-banking. Research limitations/implications The actual variance explained by our study’s model was higher in Australia (59.3%) than for Thailand (23.8%), suggesting future research of m-banking adoption in emerging Asian cultures. Practical implications We identify the important factors consumers consider when adopting m-banking. The findings of this research give banking organisations a foundational model that can be used to support m-banking implementation. Originality/value Our study is perhaps the first to examine and compare the intention to adopt m-banking across Thai and Australian consumers, and responds to calls for additional research that generalises m-banking and m-services acceptance across cultures. This study has proposed and validated additional constructs that are not present in the original SST Intention to Use model.