Audio-visual System Research Articles

Medical Tele-Diagnosis Robot (MTR) - Internet Based Communication & Navigation System

Medical Tele-diagnosis Robot (MTR) is a cost effective telemedicine mobile robot that provides tele-presence capability for the specialist on a remote location to virtually meet the patient, perform diagnostics and consult the resident doctor regarding the patient via internet. This paper highlights on the development of a doctor-robot interface where the doctor or user can control the robot reliably via regular internet connection from a different location, a distributed secured network for MTRs communication, an audiovisual communication system for tele-diagnosis and a navigation safety system called Danger Monitoring System (DMS) as part of MTRs assistive internet based navigation remote control system. The overall setup and maintenance cost of MTR is reduced by adopting a decentralized network via hybrid P2P technology. With this, the network load is distributed among the users. As for the audiovisual system, the timeliness of the video transmission from the robot to the operator can be attained by CUDA H.264 video encoding to reduce the size of the video stream and by taking advantage of the highly-parallel processors in the graphics processing unit. Combinations of sensors are place around the robot to provide data on the robots surrounding during operation. The sensors data are fed into the DMS algorithm which is equipped with fuzzy logic based artificial intelligence system to process the data from all the sensors and user input to decide preventative measures to avoid any danger to humans and the robot in terms of obstacle avoidance and robot tilt angle safety. The overall system is tested by a set of experiments and found to be demonstrating an acceptable performance. This system proved to be suitable to be used in MTR.

Feature Fusion Based Audio-Visual Speaker Identification Using Hidden Markov Model under Different Lighting Variations

The aim of the paper is to propose a feature fusion based Audio-Visual Speaker Identification (AVSI) system with varied conditions of illumination environments. Among the different fusion strategies, feature level fusion has been used for the proposed AVSI system where Hidden Markov Model (HMM) is used for learning and classification. Since the feature set contains richer information about the raw biometric data than any other levels, integration at feature level is expected to provide better authentication results. In this paper, both Mel Frequency Cepstral Coefficients (MFCCs) and Linear Prediction Cepstral Coefficients (LPCCs) are combined to get the audio feature vectors and Active Shape Model (ASM) based appearance and shape facial features are concatenated to take the visual feature vectors. These combined audio and visual features are used for the feature-fusion. To reduce the dimension of the audio and visual feature vectors, Principal Component Analysis (PCA) method is used. The VALID audio-visual database is used to measure the performance of the proposed system where four different illumination levels of lighting conditions are considered. Experimental results focus on the significance of the proposed audio-visual speaker identification system with various combinations of audio and visual features.

About neural-network algorithms application in viseme classification problem with face video in audiovisual speech recognition systems

The paper considers the phoneme recognition by facial expressions of a speaker in voice-activated control systems. We have developed a neural network recognition algorithm by using the phonetic words decoding method and the requirement for isolated syllable pronunciation of voice commands. The paper presents the experimental results of viseme (facial and lip position corresponding to a particular phoneme) classification of Russian vowels. We show the dependence of the classification accuracy on the used classifier (multilayer feed-forward network, support vector machine, k-nearest neighbor method), image features (histogram of oriented gradients, eigenvectors, SURF local descriptors) and the type of camera (built-in or Kinect one). The best accuracy of speaker-dependent recognition is shown to be 85% for a built-in camera and 96% for Kinect depth maps when the classification is performed with the histogram of oriented gradients and the support vector machine.

Development of telerehabilitation application with designated consultation categories.

Telerehabilitation (TR) is getting ever more popular because it is effective in bringing rehabilitation services to rural populations by means of audiovisual systems and its initial implementation studies presented encouraging results. TR is proven to be helpful, with benefits in terms of reduced travel time, cost, and availability of specialists' support in local communities. However, TR systems that are usable under low-bandwidth network environments are rare. This article introduces the development of a TR system with customized consultation categories for users to choose from, depending on requirements. Each category, with its preset parameter values, is discussed in detail by demonstrating relevant rehabilitation exercises. A novel bandwidth adaptation algorithm is also presented for optimal utilization of the dynamic network conditions, which ensures the system functionality even under narrow-bandwidth environments. Experiment results show that the system is able to perform effectively in each consultation category while the rehabilitation exercises are being performed. The proposed algorithm is also verified for its ability to adapt the content quality and effectively utilize the network under constrained conditions. A survey conducted on the video quality of the system under low-bandwidth conditions shows encouraging results for a large scale deployment of the application.

Brain areas activated by uncertain reward-based decision-making in healthy volunteers.

Reward-based decision-making has been found to activate several brain areas, including the ventrolateral prefrontal lobe, orbitofrontal cortex, anterior cingulate cortex, ventral striatum, and mesolimbic dopaminergic system. In this study, we observed brain areas activated under three degrees of uncertainty in a reward-based decision-making task (certain, risky, and ambiguous). The tasks were presented using a brain function audiovisual stimulation system. We conducted brain scans of 15 healthy volunteers using a 3.0T magnetic resonance scanner. We used SPM8 to analyze the location and intensity of activation during the reward-based decision-making task, with respect to the three conditions. We found that the orbitofrontal cortex was activated in the certain reward condition, while the prefrontal cortex, precentral gyrus, occipital visual cortex, inferior parietal lobe, cerebellar posterior lobe, middle temporal gyrus, inferior temporal gyrus, limbic lobe, and midbrain were activated during the ‘risk’ condition. The prefrontal cortex, temporal pole, inferior temporal gyrus, occipital visual cortex, and cerebellar posterior lobe were activated during ambiguous decision-making. The ventrolateral prefrontal lobe, frontal pole of the prefrontal lobe, orbitofrontal cortex, precentral gyrus, inferior temporal gyrus, fusiform gyrus, supramarginal gyrus, inferior parietal lobule, and cerebellar posterior lobe exhibited greater activation in the ‘risk’ than in the ‘certain’ condition (P < 0.05). The frontal pole and dorsolateral region of the prefrontal lobe, as well as the cerebellar posterior lobe, showed significantly greater activation in the ‘ambiguous’ condition compared to the ‘risk’ condition (P < 0.05). The prefrontal lobe, occipital lobe, parietal lobe, temporal lobe, limbic lobe, midbrain, and posterior lobe of the cerebellum were activated during decision-making about uncertain rewards. Thus, we observed different levels and regions of activation for different types of reward processing during decision-making. Specifically, when the degree of reward uncertainty increased, the number of activated brain areas increased, including greater activation of brain areas associated with loss.

Board 330 - Research Abstract Simulation-Based Assessment of ECMO Clinical Specialists (Submission #442)

Introduction/Background Extracorporeal Membrane Oxygenation [ECMO] is a life-saving measure with significant risks of morbidity and mortality for critically ill patients in cardiac or pulmonary failure. Clinical specialists in the management of ECMO patients provide continuous bedside monitoring and are the first line of defense in recognizing and responding to acute and potentially catastrophic events. However, since such events are infrequent and unpredictable, an ECMO specialist with many years of experience may never have encountered one. The purpose of this study was to create simulated scenarios of critical events to enhance the training of ECMO specialists. A second goal was to assess their performance in recognizing and responding to the simulated events. The hypothesis was that more experienced individuals would outperform the less experienced. The simulations provided a hands-on, real-time approach to the challenging and complex nature of bedside ECMO emergencies. Methods Eight scenarios, designed to reflect acute events an ECMO specialist would be expected to manage included hyperventilation, air entrainment, ventricular tachycardia, hypovolemia causing venous occlusion, agitation with arterial occlusion, pneumothorax, bleeding and hemodilution. The scenarios were created and piloted by perfusionists, critical care physicians and the ECMO nurse coordinator. These scenarios represent events that have occurred while a patient was on ECMO support. Twenty five ECMO clinical specialists consented to participate in this IRB-approved study. The subjects were oriented to the manikin and the study format and participated individually. They had five minutes to complete each scenario and an in-depth debriefing was held at the Conclusion of the session. The sessions were captured on an audiovisual system. Two trained raters scored for completion of key actions and assessed a global score. Results All twenty five participants completed the scenarios. Thirteen (52%) had less than one year of experience on the ECMO team. The generalizability coefficients were moderate for both the checklist (0.41) and global scores (0.43). Large proportions of variance were attributable to the Person x Task (P x T) interaction, suggesting that performance in one scenario was poorly predicted by performance in another. Relatively little variance was attributable to the rater, or associated interactions, indicating the raters were reasonably consistent in the assignment of scores. The standardized inter-rater reliability coefficient were comparable for the checklist scores (r=0.72) and the global scores (r=0.70). For the ANOVA of both the checklist and global scoring, the group (experience) by scenario interaction was not significant indicating that any performance differences between ECMO groups were reasonably consistent across scenarios. There was a main effect attributable to group by checklist (F=4.3, p<0.05) and global score (F=8.4, p<0.015) indicating that the more experienced ECMO group outperformed the less experienced group. Participants with >1 year experience (mean=77.5, SD=17.9) outperformed those with <=1 year experience (mean=73.1, SD=14.1) by checklist scoring as well as in their global score [>1 year: (mean=7.2, SD=1.5) versus <=1 year: (mean=6.6, SD=1.4)]. There was also a main effect attributable to scenario based on checklist scores (F=6.9, p<0.01), as well as the global rating (F=4.1, p<0.01) indicating that the scenarios were not of equal difficulty. Conclusion ECMO simulation scenarios were created for acute skills training and evaluation of ECMO clinical specialists. Regarding reliability, the study revealed a great deal of task specificity, which is not unexpected with a heterogeneous group of tasks to be accomplished. Construct validity was supported by the experienced specialists outperforming the less experienced. It was also supported by participant feedback following the sessions. In their evaluations, the participants felt that the simulation scenarios were realistic, represented events that could occur in practice and were appropriate for training. These preliminary data support the simulation of ECMO critical events for training and evaluation of clinical specialists.

Board 173 - Program Innovations Abstract Transfer into the Practice of the ACRM Principles after a Clinical Simulation Course (Submission #1317)

Introduction/Background Currently, the formation of anesthesia and CPR specialists considers as part of its curriculum the acquisition of attitudinal competences developing skills in: communication, leadership, decision making, teamwork and conflict resolution in crisis situations, which may reduce the occurrence of adverse outcomes and legal problems, in addition to shortening the learning curves in acquiring these skills. The present work describes the development of the first ACRM course in Chile, based in high fidelity clinical simulation for anesthesia residents at Diego Portales University. The objective is to describe the development of the first ACRM course in Chile and evaluate the teaching quality provided and the perceived non-technical skills attainment. Methods A course was organized into two, five hour blocks for 18 first and second year anesthesia residents divided into five groups. Prior to the beginning of the course, each of the participants received documents for self-study about ACRM. Each activity block was formally carried out with high fidelity simulation and debriefing for four participants and it considered four scenarios validated by the University of Western Ontario. The assessment of the simulation day was performed using the Debriefing Assessment Simulation in Healthcare ("DASH"®) instrument, in its Spanish version for students. Achieved long term non-technical skills assessment was performed at eight months post-course using the Spanish version of the "Anesthetics Non-Technical Skills" (ANTS) instrument. The Spanish version of the ANTS and DASH instruments, were developed by the researchers who conducted the process of semantic validation using the Delphi methodology. The analysis was performed using the SPSS18 software. All of the participants had a 100% attendance to the scheduled activities and performed scenarios recommended by the American Society of Anesthesia (ASA) for ACRM courses. The DASH assessment showed a high valuation in all elements and descriptors (average grade 6.7), with a lower relative valuation related to the use of audiovisual systems (average grade 5.6). The ANTS evaluation was answered by 15 out of the 18 participants (error rate of 10.6% for 95% of IC). The categories with the best improvement perception correspond to decision making and task handling (77% and 81% of the students, respectively). In the teamwork field, there is a good observation in 69% of the students, with significant differences in the valuation of the different elements that form this category. The situational awareness item was the one that showed the least improvement attained with a 56%. The evaluation showed an improvement perception with regards to knowledge and handling of the situation when they faced real life situations in daily practice. Results: Conclusion Incorporating ACRM courses in initial training levels for the formation of anesthesia specialists, provides major tools for the development of non-technical skills in these professionals, within a safe and positive environment according to the valuations obtained using the DASH instrument. The improving perception is high, however, probably due to the lack of experience and with the need of strengthening their own knowledge of the specialty, advanced training courses may have a greater impact on the participants. This is the first experience we have had in Chile with this type of course in anesthesia residents.

Information Seeking Behavior Habits of Teachers and Scholars for Use of E-Resources in Engineering College Libraries – A Survey

The study aims to get the information seeking behavior habits of teachers and scholars for use of E-Resources in engineering college libraries of Bangalore region under Visvesvaraya Technological University, Belgaum,Karnataka. The internet and web are constantly influencing the development of scholarly communication work. The teachers and scholars were considered as core group in using the resources to the maximum extent. Ultimately, sufficient computers, adequate facility of internet, audio visual systems,multimedia equipments and e-book readers are considered as tool to access the e-resources at their liberty. The way the developments have taken place in the academic environment particularly in engineering college libraries has changed the Librarian role. This paper will discuss about the facility, technical feasibility and ensures the users opinion about the library staff involvement and co-operation.

The Semiotic Status of Interlingual Subtitling

This article examines the question of the semiotic status of interlingual subtitling. More specifically, it will be shown that the conception of subtitling as intersemiotic translation is theoretically untenable and therefore cannot provide the basis for an adequate approach to the peculiar character of this type of audiovisual translation. The semiotic status of interlingual subtitling is essentially determined by the particular nature of the audiovisual system. Thus, a semiotic description of the audiovisual signifying processes will be offered that will allow for a conception of the semiotic status of interlingual subtitling in intertextual terms.

G214 The Best Way to Teach Developmental Assessment – a Single Blinded Randomised Study Comparing Teaching Models

AimsDevelopmental assessment is a core paediatric competency but research demonstrates teaching gaps. This single blinded, randomised controlled study compares three teaching approaches to developmental assessment in a large group setting...

SU-E-T-358: Double-Ends Quasi-Breath-Hold (DE-QBH) Technique for Respiratory Motion Management

Purpose: To introduce double‐ends quasi‐breath‐hold (DE‐QBH) technique and evaluate its feasibility. Methods: It was previously demonstrated QBH, a hybrid technique combining free‐breathing‐gating (FBG) and breath‐hold, could provide less motion uncertainty and shorter treatment time than conventional FBG. While QBH uses only one phase, either end‐of‐exhalation (EOE) or end‐of‐inhalation (EOI), DE‐QBH utilizes both phases to further improve delivery efficiency. DE‐QBH is realized using an audiovisual biofeedback system (AVBFS) and a respiratory motion management program (RMMP). The in‐house developed AVBFS, consisting of two infrared stereo cameras and a head mounted display, provides dynamic feedback to patients. The RMMP establishes a personalized DE‐QBH model by adding a short quasi‐breath‐hold period to both EOI and EOE phase. Then the patient is guided to follow the model. A simulation study for 6 different maneuvers (2−2, 3−3, 4−4, 5−5, 7−5, and 9−5 sec for EOI‐EOE combination) was performed with 3 volunteers. External motion signals were analyzed to obtain mean absolute error (MAE) between the measured and guided. Duty cycle was also estimated. Results: MAEs were smaller than 1 mm for all maneuvers except 9−5 combination [0.9(+/−)1.3, 0.7(+/−)1.2, 0.6(+/−)1.1, 0.7(+/−)1.1, 0.6(+/−)1.0, and 1.1(+/−)0.8 mm for 2−2, 3−3, 4−4, 5−5, 7−5, and 9−5 combination, respectively]. Estimated duty cycles were 43, 52, 57, 62, 64, and 71% for 2−2, 3−3, 4−4, 5−5, 7−5, and 9v5 combination, respectively. Conclusion: The proposed DE‐QBH technique was feasible for respiratory motion management. It showed excellent feasibility with minimal motion uncertainty and significantly improved delivery efficiency, reaching up to 70% duty cycle.

La formation des enseignants recourant aux dispositifs audio-visuels : analyse des configurations d’activité chez des fonctionnaires stagiaires

Cet article a pour objectif, à travers une étude de cas, de présenter les modes d’appropriation d’un dispositif de formation des enseignants ayant recours au support audio-visuel. L’étude envisage les situations de formation intégrant un dispositif audio-visuel comme des situations d’interaction entre un individu, un collectif et un artefact. L’analyse porte sur les configurations d’activités produites par les formés c’est-à-dire les formes auto-organisées d’interaction individu-environnement social et matériel. Les résultats révèlent différents types de médiation instrumentale développés par les formés, offrant des aides au développement professionnel.

Enhancing audiovisual experience with haptic feedback: a survey on HAV.

Haptic technology has been widely employed in applications ranging from teleoperation and medical simulation to art and design, including entertainment, flight simulation, and virtual reality. Today there is a growing interest among researchers in integrating haptic feedback into audiovisual systems. A new medium emerges from this effort: haptic-audiovisual (HAV) content. This paper presents the techniques, formalisms, and key results pertinent to this medium. We first review the three main stages of the HAV workflow: the production, distribution, and rendering of haptic effects. We then highlight the pressing necessity for evaluation techniques in this context and discuss the key challenges in the field. By building on existing technologies and tackling the specific challenges of the enhancement of audiovisual experience with haptics, we believe the field presents exciting research perspectives whose financial and societal stakes are significant.

Seamless integration of audio visual design into architecture for more successful projects

The integration of audio visual systems into technology-enabled buildings has long been a struggle with the architectural community. In some instances, architects do not want any part of the systems to be visible within the space, like speakers, projectors and equipment racks. In other cases, like digital signage and way-finding technology, the systems are front and center in the room and must be as visible and as user-accessible as possible. Audio visual technology is present in some form in almost every building being constructed today including educational, corporate, houses of worship, legislative, hospitality, sports arenas, healthcare, museums, and retail. Much time is spent not just on the system design, but working with architects on integrating the technology into these built environments, each of which has its own requirements. This presentation focuses on various ways to integrate audio visual technology into buildings for more seamless and successful projects. Balancing the functionality and performance of any audio visual system installation with the aesthetic impact on the space takes experience, creativeness, and a willingness to coordinate with the design team.

Multi-pose lipreading and audio-visual speech recognition

In this article, we study the adaptation of visual and audio-visual speech recognition systems to non-ideal visual conditions. We focus on overcoming the effects of a changing pose of the speaker, a problem encountered in natural situations where the speaker moves freely and does not keep a frontal pose with relation to the camera. To handle these situations, we introduce a pose normalization block in a standard system and generate virtual frontal views from non-frontal images. The proposed method is inspired by pose-invariant face recognition and relies on linear regression to find an approximate mapping between images from different poses. We integrate the proposed pose normalization block at different stages of the speech recognition system and quantify the loss of performance related to pose changes and pose normalization techniques. In audio-visual experiments we also analyze the integration of the audio and visual streams. We show that an audio-visual system should account for non-frontal poses and normalization techniques in terms of the weight assigned to the visual stream in the classifier.

Abstract 3531: Clinical Outcomes Following Endovascular Stroke Treatment Facilitated via Telestroke Evaluation

BACKGROUND: Most stroke patients present to small community hospitals without established stroke pathways or interventional stroke treatment capability. The advent of 2 way audiovisual telestroke systems gives such patients the opportunity to be assessed rapidly by stroke neurologists. Patients who are not candidates for systemic IV tPA or have failed thrombolytic treatment can be identified and transferred to a comprehensive stroke center for potential endovascular treatment. We compared the clinical outcomes of patients undergoing endovascular stroke treatment at University of Pittsburgh Medical Center triaged either through telestroke or non-telestroke means. METHODS: Prospective data including demographics, co-morbidities, baseline Alberta Stroke Program Early CT (ASPECT) and National Institute of Health Stroke Scale (NIHSS) score, intervention modality (pharmacological, mechanical or both), time to treatment, clinical outcome, and hemorrhage and mortality rates were compared. Favorable outcome was defined as modified rankin score (mRS) of 2 or less. RESULTS: Between 3/2007 and 5/2011, thirty four patients underwent endovascular stroke treatment following telestroke evaluation versus 354 patients who were triaged through other means. Baseline characteristics were similar between the groups. Time to endovascular treatment (595 vs. 767 minutes; p = 0.5), pretreatment with systemic tPA (51.6 vs. 56.9%, p=0.6), recanalization (TIMI ≥ 2; 91.2% vs. 84.8%; p = 0.31), favorable outcome (modified rankin score ≤ 2; 50% vs. 40.4%; p = 0.29) and mortality rates (28.1% vs. 34.9%, p=0.44) were comparable. Multivariate logistic regression model identified young age (OR 0.91, CI 0.88-0.95, p&lt;0.01), successful recanalization (OR 3.3, CI 1.8-6.2, p&lt;0.01), and baseline ASPECT score (OR 6.5, CI 2.4-17.4, p&lt;0.01) as predictors of favorable outcome. CONCLUSION: The results of this study suggest that telestroke guided endovascular stroke treatment is feasible and the outcomes are similar to those patients who were triaged by traditional means. Future randomized studies which specifically compare triage via telemedicine vs. telephone or direct emergency department presentation are needed to substantiate these findings.

Patient Adaptive Neurofeedback for ADHD Therapy

An EEG-Based audiovisual stimulation system was tested on a group of 32 ADHD patients. The efficacy was compared in a pre- and post analysis over a treatment period of 3 month with a control group and a group of patients with an EEGbased neurofeedback therapy. The number of impulsivity errors in a stop-signal paradigm at several distinct times during and after the therapy was used as a quality marker. The main feature of this method uses real-time-EEG data derived parameters for a controlled manipulation of EEG patterns with persistently strong audio-visual stimuli. It is based upon a direct feedback of changes in short-term coherences in the context of central nervous stimulus processing to control the stimulating frequency. The real-time EEG processing is done on the basis of time-dependent coherence analysis of the theta and beta band.

The effectiveness of interventions to reduce anxiety, claustrophobia, sedation and non-completion rates of patients undergoing high technology medical imaging.

Advanced, high technology medical imaging, such as Computed Tomography (CT), Magnetic Resonance Imaging (MRI), Positron Emission Tomography (PET) and Single Photon Emission Computed Tomography (SPECT), have seen significant increases in their use. A number of interventions have been developed with the aim of reducing anxiety, fear and claustrophobia prior to and during imaging. To determine the effectiveness of interventions aiming to improve the patient experience of high technology medical imaging. Any patients who have undergone high technology (MRI, CT, PET, Nuclear medicine and SPECT) imaging in a medical imaging department. Interventional scans were not considered for inclusion.Considered interventions were: MRI scanner designs, quieter machines, information, psychological support, anxiety reduction protocols, mock MRI, team training, prone scanning, audiovisual (AV) systems, cognitive behavioural therapies, fragrance administration, guided imagery, patient positioning devices and paediatric preparation booklets.Experimental and observational studies were included in the review.Primary patient outcomes:Secondary patient outcome measures included: SEARCH STRATEGY: A comprehensive, three-step search strategy conducted during August to October 2011 aimed to find both published and unpublished English language studies across 14 major databases. No time restrictions. Retrieved papers were assessed by two independent reviewers prior to inclusion in the review using standardised critical appraisal instruments from the Joanna Briggs Institute Meta Analysis of Statistics Assessment and Review Instrument (JBI-MAStARI). Data was extracted from papers included in the review using the standardised data extraction tool from JBI-MAStARI. Quantitative data was, where possible pooled in statistical meta-analysis using JBI-MAStARI. All results were subject to double data entry. Thirty-eight studies met the inclusion criteria and were considered to be of suitable methodological quality. The quality of the studies varied; however, all the studies included had at a minimum a comparison group. Of the final 38 studies, 29 assessed interventions for MRI, 8 Nuclear Medicine procedures, and one in CT. Meta-analysis was performed for two interventions; additional information and an AV system.Significantly fewer children (10-18 years) required sedation prior to a MRI scan when given an audiovisual intervention (n=1785) OR 0.42% (CI 0.25 - 0.70), compared with the control group. Open MRI, newer MRI scanners, quieter machines, information in combination with psychological support or an anxiety reduction protocol, mock MRI, team training, prone scanning, AV systems, cognitive behavioural therapies, fragrance administration, guided imagery, patient positioning devices and paediatric preparation booklets were all found to have some positive effect on at least one outcome, whilst the findings for additional information were mixed. This review identified 38 studies that assessed interventions to reduce anxiety, claustrophobia, sedation and non-completions and improve satisfaction for patients undergoing medical imaging. The majority of the interventions studied had some positive effect on at least one outcome.Healthcare professionals need to be aware of the potential for patients undergoing medical imaging to experience anxiety, fear and claustrophobia during scanning. A number of interventions can be selected and tailored to the patient group.There is still a significant amount of research required to establish the effectiveness for a number of interventions aimed to reduce anxiety, fear and claustrophobia in medical imaging, particularly in CT.

Multi-pose lipreading and audio-visual speech recognition

In this article, we study the adaptation of visual and audio-visual speech recognition systems to non-ideal visual conditions. We focus on overcoming the effects of a changing pose of the speaker, a problem encountered in natural situations where the speaker moves freely and does not keep a frontal pose with relation to the camera. To handle these situations, we introduce a pose normalization block in a standard system and generate virtual frontal views from non-frontal images. The proposed method is inspired by pose-invariant face recognition and relies on linear regression to find an approximate mapping between images from different poses. We integrate the proposed pose normalization block at different stages of the speech recognition system and quantify the loss of performance related to pose changes and pose normalization techniques. In audio-visual experiments we also analyze the integration of the audio and visual streams. We show that an audio-visual system should account for non-frontal poses and normalization techniques in terms of the weight assigned to the visual stream in the classifier.

Intensive Care Unit Telemedicine

1Telemedicine services used in the intensive care setting today, which provide continuous monitoring to hundreds of patients across multiple sites, have been in use since the year 2000. The term tele-ICU is now used to describe a concept of care in which a centralized or remotely based critical care team is networked with the bed side intensive care unit (ICU) team and patient via state-ofthe-art audiovisual communication and computer systems. 2