This letter shows by digital simulation that a simple rule applied to one-dimensional self-organized maps for integrating sensory perceptions from two identical sources yielding position information as integers, corrupted by independent noise sources, yields almost statistically optimal results for position estimation as determined by maximum likelihood estimation. There is no learning of the corrupting noise sources nor is any information about the statistics of the noise sources available to the integrating process. The simple rule employed yields a measure of the quality of the estimated position of the source. The letter also shows that if the Bayesian estimates, which are rational numbers, are rounded in order to comply with the stipulation that integers be identified, the Bayesian estimation will have a larger variance than the proposed integration.

A neural network architecture, subjected to incon-gruent stimuli in the form of lip reading of spoken syllables and listening to different spoken syllables, is shown to generate the well-known McGurk effect, e.g. visual /ga/ and auditory /ba/ is perceived as /da/ by the network. The neural network is based on an architecture which has previously been successfully applied to sensory integration of congruent stimuli and is here extended to take into account that lip reading groups consonants into equivalence classes, bilabial, dento-labial and nonlabial consonants, rather than distinguishing between individual consonants.

Intelligent Industrial Processes (IIP) and Enabling Information and Communication Technology (Enabling ICT) are two out of the nine areas of excellence in research and innovation at the Luleå University of Technology (LTU), which are formed to foster interdisciplinary research and innovation in strategically important areas. This report presents a perspective on the role of machine learning and intelligence in these two areas, focusing in particular on future ICT for industrial process automation (ProcessIT) up to the year 2030. The study that is presented here complements similar studies made in other fields, with the common goal to create the first inputs for a broader discussion and formulation of strategic objectives in the form of a roadmap.This report presents my interpretation of the concept of Intelligent Industrial Processes and the role of ICT in that context, including novel information processing methods and devices that are inspired by biological circuits and systems. This report also includes brief introductions and definitions of important concepts; a summary of seven documents presenting international strategic agendas and objectives; a summary of identified strengths, weaknesses, opportunities and threats; a description of selected research trends with references to interesting results; a tentative outline of interesting research problems and first steps towards 2030; and a list of research groups with complementary competences that may merit future partnership. It is concluded that the Open Research and Innovation Platform that is outlined in a parallel study would be a valuable resource for machine-learning research, development and education because transparent access to data is a key enabling factor. In terms of machine-learning research it is concluded that we need to take the step from studies of isolated learning algorithms and applications to closed-loop learning architectures for large-scale sensor-actuator systems, possibly including human- machine interaction, decision support systems and models of complex systems such as maintenance systems and markets.The aim to develop intelligent industrial processes using a new generation of ICT is an ambitious interdisciplinary initiative, which is likely to force us thinking beyond conventional methods and to educate a new generation of engineers that understand the necessary concepts.

We present a recurrent multimodal model of binding written words to mental objects and investigate the capability of the network in reading misspelt but categorically related words. Our model consists of three mutually interconnected association modules which store mental objects, represent their written names and bind these together to form mental concepts. A feedback gain controlling top-down influence is incorporated into the model architecture and it is shown that correct settings for this during map formation and simulated reading experiments is necessary for correct interpretation and semantic binding of the written words.

The multimodal self-organizing network (MMSON), an artificial neural network architecture carrying out sensory integration, is presented here. The architecture is designed using neurophysiological findings and imaging studies that pertain to sensory integration and consists of interconnected lattices of artificial neurons. In this artificial neural architecture, the degree of recognition of stimuli, that is, the perceived reliability of stimuli in the various subnetworks, is included in the computation. The MMSON's behavior is compared to aspects of brain function that deal with sensory integration. According to human behavioral studies, integration of signals from sensory receptors of different modalities enhances perception of objects and events and also reduces time to detection. In neocortex, integration takes place in bimodal and multimodal association areas and result, not only in feedback-mediated enhanced unimodal perception and shortened reaction time, but also in robust bimodal or multimodal percepts. Simulation data from the presented artificial neural network architecture show that it replicates these important psychological and neuroscientific characteristics of sensory integration.

Towards hardware implementation of real-time visual image processing, we propose a region-based coupled Markov Random Field (MRF) model with phases as hidden variables for coarse image region segmentation tasks. In general, two types of coupled MRF models, boundary-based and region-based, are known according to their hidden variables that play a crucial role for detecting discontinuities in intensity, color, depth and motion in image scenes. A region-based coupled MRF model with phases as hidden variables has been proposed for image restoration task. For coarse region image segmentation tasks, we customize the previous model in view of effient hardware implementation. Our model has an advantage over the resistive-fuse network, which is a boundary-based coupled MRF model, in dealing with the hidden variables explicitly. Consequently, our model can extract closed regions from given images by phases as labels at di®erent timing and act as a spatial-temporal nonlinear filter.

Integration of signals from sensory receptors of different modalities is known to enhance perception. Integration takes place in bimodal and multimodal association areas of neocortex and results in robust bimodal or multimodal percepts as well as in feedback mediated enhanced unimodal perception. A Multimodal Self-Organizing Network, Mu-SON, is presented as a tool for simulating sensory integration. The latest version of this MuSON, that is described in the current paper, also takes the degree of recognition of stimuli in the various maps of the network into account. Simulation results show the same characteristics as corresponding results from psychology and neuroscience.

We present a model of integration of auditory and visual information as in the human cortex. More specifically, we demonstrate a possible way in which the phonetic symbols and associated Mandarin Chinese phonemes pronounced by different speakers are mapped onto the model of cortical areas. Our model has been strongly influenced by recent fMRI studies on integration of letters and speech sounds in the human brain. The model is based on multimodal self-organizing networks (MuSoNs) which were introduced in our previous works and proved to be a convenient tool to describe and study mapping and integration of sensory information as in the cortex. The model also shows how phonemes pronounced by different speakers are mapped onto overlapping cortical areas

Multimodal integration of sensory information has clear advantages for survival: events that can be sensed in more than one modality are detected more quickly and accurately, and if the sensory information is corrupted by noise the classification of the event is more robust in multimodal percepts than in the unisensory information. It is shown that using a Multimodal Self-Organizing Network (MuSON), consisting of several interconnected Kohonen Self-Organizing Maps (SOM), bimodal integration of phonemes, auditory elements of language, and letters, visual elements of language, can be simulated. Robustness of the bimodal percepts against noise in both the auditory and visual modalities is clearly demonstrated.

It is known from psychology and neuroscience that multimodal integration of sensory information enhances the perception of stimuli that are corrupted in one or more modalities. A prominent example of this is that auditory perception of speech is enhanced when speech is bimodal, i.e. when it also has a visual modality. The function of the cortical network processing speech in auditory and visual cortices and in multimodal association areas, is modeled with a Multimodal Self-Organizing Network (MuSON), consisting of several Kohonen Self-Organizing Maps (SOM) with both feedforward and feedback connections. Simulations with heavily corrupted phonemes and uncorrupted letters as inputs to the MuSON demonstrate a strongly enhanced auditory perception. This is explained by feedback from the bimodal area into the auditory stream, as in cortical processing.