Efficient SLAM algorithm with hybrid visual map in an indoor environment

Exploiting the relevant speech information that is embedded in facial images has been a significant research topic in recent years, because it has provided complementary information to acoustic signals for a wide range of automatic speech recognition (ASR) tasks. Visual information is particularly important in many real applications where acoustic signals are corrupted by environmental noises. This chapter reviews the most recent advances in feature extraction and representation for Visual Speech Recognition (VSR). In comparison with other surveys published in the past decade, this chapter presents a more up-to-date survey and highlights the strengths of two newly developed approaches (i.e., graph-based learning and deep learning) for VSR. In particular, we summarise the methods of using these two techniques to overcome one of the most challenging difficulties in this area-that is, how to automatically learn good visual feature representations from facial images to replace the widely used handcrafted features. This chapter concludes by discussing potential visual feature representation solutions that may overcome the remaining challenges in this domain.

Series feature aggregation for content-based image retrieval

The ever-increasing numbers of medical images in digital format are generated in clinical practice every day. These images include high-resolution and temporal data of various modalities and which constitute an important source of anatomical and functional information for diagnosis of diseases, research and education. So how to manage these large data and utilize them effectively and efficiently possess significant technical challenges. Thus, the technique of Content-based Medical Image Retrieval (CBMIR) emerges as the times require. However, current CBMIR is not sufficient to capture the semantic content of an image and difficult to provide good results according to the predefined categories in the medical domain for less using the medical knowledge. Accordingly, in this paper a mapping modelling of visual feature and knowledge representation is proposed to approach for medical image retrieval. Firstly, the low-level fusion visual features are extracted based on intensity, texture, and their extended versions. Secondly, a set of disjoint semantic tokens with appearance in lung CT images is selected to define a vocabulary based on medical knowledge representation. Finally, a mapping modeling is investigated to associate low-level visual image features with their high-level semantic. Experiments are conducted with a medical image database consisting of 220 lung medical images obtained from the Heilongjiang Province Hospital. And the comparison of the retrieval results shows that the approach proposed in this paper is effective.

We propose associating language utterances to 3D visual abstractions of the scene they describe. The 3D visual abstractions are encoded as 3-dimensional visual feature maps. We infer these 3D visual scene feature maps from RGB images of the scene via view prediction: when the generated 3D scene feature map is neurally projected from a camera viewpoint, it should match the corresponding RGB image. We present generative models that condition on the dependency tree of an utterance and generate a corresponding visual 3D feature map as well as reason about its plausibility, and detector models that condition on both the dependency tree of an utterance and a related image and localize the object referents in the 3D feature map inferred from the image. Our model outperforms models of language and vision that associate language with 2D CNN activations or 2D images by a large margin in a variety of tasks, such as, classifying plausibility of utterances, detecting referential expressions, and supplying rewards for trajectory optimization of object placement policies from language instructions. We perform numerous ablations and show the improved performance of our detectors is due to its better generalization across camera viewpoints and lack of object interferences in the inferred 3D feature space, and the improved performance of our generators is due to their ability to spatially reason about objects and their configurations in 3D when mapping from language to scenes.

This thesis presents a recent research on the problem of environmental modeling for both localization and map building for wheel-based, differential driven, fully autonomous and self-contained mobile robots. The robots behave in an indoor office environment. They have a multi-sensor setup where the encoders are used for odometry and two exteroperceptive sensors, a 360° laser scanner and a monocular vision system, are employed to perceive the surrounding. The whole approach is feature based meaning that instead of directly using the raw data from the sensor features are firstly extracted. This allows the filtering of noise from the sensors and permits taking account of the dynamics in the environment. Furthermore, a properly chosen feature extraction has the characteristic of better isolating informative patterns. When describing these features care has to be taken that the uncertainty from the measurements is taken into account. The representation of the environment is crucial for mobile robot navigation. The model defines which perception capabilities are required and also which navigation technique is allowed to be used. The presented environmental model is both metric and topological. By coherently combining the two paradigms the advantages of both methods are added in order to face the drawbacks of a single approach. The capabilities of the hybrid approach are exploited to model an indoor office environment where metric information is used locally in structures (rooms, offices), which are naturally defined by the environment itself while the topology of the whole environment is resumed separately thus avoiding the need of global metric consistency. The hybrid model permits the use of two different and complementary approaches for localization, map building and planning. This combination permits the grouping of all the characteristics which enables the following goals to be met: Precision, robustness and practicability. Metric approaches are, per definition, precise. The use of an Extended Kalman Filter (EKF) permits to have a precision which is just bounded by the quality of the sensor data. Topological approaches can easily handle large environments because they do not heavily rely on dead reckoning. Global consistency can, therefore, be maintained for large environments. Consistent mapping, which handle large environments, is achieved by choosing a topological localization approach, based on a Partially Observable Markov Decision Process (POMDP), which is extended to simultaneous localization and map building. The theory can be mathematically proven by making some assumptions. However, as stated during the whole work, at the end the robot itself has to show how good the theory is when used in the real world. For this extensive experimentation for a total of more than 9 km is performed with fully autonomous self-contained robots. These experiments are then carefully analyzed. With the metric approach precision with error bounds of about 1 cm and less than 1 degree is further confirmed by ground truth measurements with a mean error of less than 1 cm. The topological approach is successfully tested by simultaneous localization and map building where the automatically created maps turned out to work better than the a priori maps. Relocation and closing the loop are also successfully tested.

Vision-language constraint graph representation learning for unsupervised vehicle re-identification

We present a novel approach to real-time dense visual simultaneous localisation and mapping. Our system is capable of capturing comprehensive dense globally consistent surfel-based maps of room scale environments and beyond explored using an RGB-D camera in an incremental online fashion, without pose graph optimization or any post-processing steps. This is accomplished by using dense frame-to-model camera tracking and windowed surfel-based fusion coupled with frequent model refinement through non-rigid surface deformations. Our approach applies local model-to-model surface loop closure optimizations as often as possible to stay close to the mode of the map distribution, while utilizing global loop closure to recover from arbitrary drift and maintain global consistency. In the spirit of improving map quality as well as tracking accuracy and robustness, we furthermore explore a novel approach to real-time discrete light source detection. This technique is capable of detecting numerous light sources in indoor environments in real-time as a user handheld camera explores the scene. Absolutely no prior information about the scene or number of light sources is required. By making a small set of simple assumptions about the appearance properties of the scene our method can incrementally estimate both the quantity and location of multiple light sources in the environment in an online fashion. Our results demonstrate that our technique functions well in many different environments and lighting configurations. We show that this enables (a) more realistic augmented reality rendering; (b) a richer understanding of the scene beyond pure geometry and; (c) more accurate and robust photometric tracking.

Recent vision-language understanding approaches adopt a multi-modal transformer pre-training and finetuning paradigm. Prior work learns representations of text tokens and visual features with cross-attention mechanisms and captures the alignment solely based on indirect signals. In this work, we propose to enhance the alignment mechanism by incorporating image scene graph structures as the bridge between the two modalities, and learning with new contrastive objectives. In our preliminary study on the challenging compositional visual question answering task, we show the proposed approach achieves improved results, demonstrating potentials to enhance vision-language understanding.

Most of the mapping methods in indoor environments use single type of feature. Typically an expert decides which feature type to be used. In this work, the robot selects suitable feature type based on the current place during mapping. During the training phase, an expert labels places with the suitable feature type. The robot calculates a metric for each feature type and trains a hidden Markov model with the labels provided by the expert. In a new environment, the robot uses the trained hidden Markov model and selects suitable feature type. Hierarchical topological mapping enables usage of different feature types in the map representation. This method increases the adaptability of the mapping system in new environments and enables different feature types to be utilized in the same environment. The devised method is applied to a robot equipped with a laser sensor and a camera in a simulation environment and tested in an indoor environment.

It's well known that visual information such as lip shape and its movement can indicate what the speaker is talking about. In this paper, we present an automatic lipreading system solely using visual information for recognizing isolated English digits from 0 to 9. A parameter set of a 14-point ASM lip model is used to describe the outer lip contour. The inner mouth information such as the teeth region and the mouth opening are also extracted. With appropriate normalization, the feature vectors containing the normalized outer lip features, inner mouth features and also their first order derivatives are obtained for training the HMM word models. Experiments have been carried out to investigate the recognition performance using our visual feature set compared with other traditional visual feature representations. An accuracy of 93% for speaker dependent recognition and 84% for speaker independent recognition is achieved using our visual feature representation. A real-time automatic lipreading system has been successfully implemented on a 1.9-GHz PC.

Interactive Machine Learning considers that a robot is learning with and/or from a human. In this paper, we investigate the impact of human social traits on the robot learning. We explore social traits such as age (children vs. adult) and pathology (typical developing children vs. children with autistic spectrum disorders). In particular, we consider learning to recognize both postures and identity of a human partner. A human-robot posture imitation learning, based on a neural network architecture, is used to develop a multi-task learning framework. This architecture exploits three learning levels : 1) visual feature representation, 2) posture classification and 3) human partner identification. During the experiment the robot interacts with children with autism spectrum disorders (ASD), typical developing children (TD) and healthy adults. Previous works assessed the impact on learning of these social traits at the group level. In this paper, we focus on the analysis of individuals separately. The results show that the robot is impacted by the social traits of these different groups' individuals. First, the architecture needs to learn more visual features when interacting with a child with ASD (compared to a TD child) or with a TD child (compared to an adult). However, this surplus in the number of neurons helped the robot to improve the TD children's posture recognition but not that of children with ASD. Second, preliminary results show that this need of a neurons surplus while interacting with children with ASD is also generalizable to the identity recognition task.

The brain represents visual objects with topographic cortical patterns. To address how distributed visual representations enable object categorization, we established predictive encoding models based on a deep residual network, and trained them to predict cortical responses to natural movies. Using this predictive model, we mapped human cortical representations to 64,000 visual objects from 80 categories with high throughput and accuracy. Such representations covered both the ventral and dorsal pathways, reflected multiple levels of object features, and preserved semantic relationships between categories. In the entire visual cortex, object representations were organized into three clusters of categories: biological objects, non-biological objects, and background scenes. In a finer scale specific to each cluster, object representations revealed sub-clusters for further categorization. Such hierarchical clustering of category representations was mostly contributed by cortical representations of object features from middle to high levels. In summary, this study demonstrates a useful computational strategy to characterize the cortical organization and representations of visual features for rapid categorization.

Combined representation of visual features in the scene-selective cortex

Indoor air quality (IAQ) parameters are not only directly related to occupational health but also have a huge impact on quality of life. In particular, besides having a very influence on the public health as it may cause a great variety of health effects such as headaches, dizziness, restlessness, difficulty breathing, increase heart rate, elevated blood pressure, coma and asphyxia, carbon dioxide (CO2) can be used as an important index of IAQ. In fact, due to people spend about 90% of our lives indoors, it is extremely important to monitor the CO2 concentration in real-time to detect problems in the IAQ in order to quickly take interventions in the building to increase the IAQ. The variation of CO2 in indoor living environments is in most situations related to the low air renewal inside buildings. CO2 levels over 1000 ppm, indicate a potential problem with indoor air. This paper aims to present iAirC a solution for CO2 real-time monitoring based on Internet of Things (IoT) architecture. This solution is composed by a hardware prototype for ambient data collection and a web and smartphone compatibility for data consulting. This system performs real-time data collection that is stored in a ThingSpeak platform and has smartphone compatibility which allows easier access to data in real time. The user can also check the latest data collected by the system and access to the history of the CO2 levels in a graphical representation. iAirC uses an open-source ESP8266 for Wi-Fi 2.4 GHZ as processing and communication unit and incorporates a CO2 sensor as sensing unit.

Much progress has been achieved during the past two decades in audio-visual automatic speech recognition (AVASR). However, challenges persist that hinder AVASR deployment in practical situations, most notably, robust and fast extraction of visual speech features. We review our efforts in overcoming this problem, based on an appearance-based visual feature representation of the speaker's mouth region. We cover three topics in particular. Firstly, we discuss AVASR in realistic, visually challenging domains, where lighting, background, and head-pose vary significantly. To enhance visual-front-end robustness in such environments, we employ an improved statistical-based face detection algorithm that significantly outperforms our baseline scheme. However, visual-only recognition remains inferior to visually clean (studio-like) data, thus demonstrating the importance of accurate mouth region extraction. We then consider a wearable audio-visual sensor to capture the mouth region directly, thus eliminating face detection. Its use improves visual-only recognition, even over full-face videos recorded in the studio-like environment. Finally, we address the speed issue in visual feature extraction, by discussing our real-time AVASR prototype implementation. The reported progress demonstrates the feasibility of practical AVASR.