Winfried A. Fellenz

Emotion recognition in human-computer interaction

Two channels have been distinguished in human interaction: one transmits explicit messages, which may be about anything or nothing; the other transmits implicit messages about the speakers themselves. Both linguistics and technology have invested enormous efforts in understanding the first, explicit channel, but the second is not as well understood. Understanding the other partys emotions is one of the key tasks associated with the second, implicit channel. To tackle that task, signal processing and analysis techniques have to be developed, while, at the same time, consolidating psychological and linguistic analyses of emotion. This article examines basic issues in those areas. It is motivated by the PKYSTA project, in which we aim to develop a hybrid system capable of using information from faces and voices to recognize peoples emotions.

On emotion recognition of faces and of speech using neural networks, fuzzy logic and the ASSESS system

We propose a framework for the processing of face image sequences and speech, using different dynamic techniques to extract appropriate features for emotion recognition. The features will be used by a hybrid classification procedure, employing neural network techniques and fuzzy logic, to accumulate the evidence for the presence of an emotional expression of the face and the speakers voice.

Establishing retinotopy by lateral-inhibition type homogeneous neural fields

Abstract We study the topographic development of visual receptive fields by simulating the continuum field equations with learning on a two-dimensional lattice. The observed plasticity reveals a columnar organization with spatial clustering of receptive field centers and the development of orientation selective neurons.

A general framework for symbol and rule extraction in neural networks

We split the rule extraction task into a subsymbolic and a symbolic phase and present a set of neural networks for filling the former. Under the two general commitments of: (i) having a learning algorithm that is sensitive to feedback signals coming from the latter phase, and (ii) extracting Boolean variables whose meaning is determined by the further symbolic processing, we introduce three unsupervised learning algorithms and show related numerical examples for a multilayer perceptron, recurrent neural networks, and a specially devised vector quantizer.

Enhancing the storage capacity of binary associative memories

We present a simple modification to binary associative memories employing a designated intermediate layer representation which reduces crosstalk between stored patterns and enhances performance and average storage capacity of the network. It is shown that the hidden layer associative network is less dependent on the number of ones in the storage patterns and shows an improved retrieval capacity from incomplete retrieval patterns.

The hidden-layer model of hippocampus

Abstract We analyze the problems facing the application to the hippocampus of a recent model of enhanced memory storage by an associative memory, achieved by insertion of a hidden layer. We extend this model to include biological constraints like the limited overall connectivity and the distributed processing in a sequence of maps. We will show that the proposed multiple layer mechanism employing a sparse code and a k -winner-take-all mechanism for the retrieval and completion of binary patterns and temporal sequences can be matched to the layers of the hippocampus, allowing some of its biological features to be understood in terms of the model.

The hidden layer associative memory model of hippocampus

An earlier model introduced by the authors (1999) for fast associative memory has shown to be an efficient solution to the storage of binary patterns and the recall from incomplete input. We extend this model to include more biologically realistic constraints to serve as a model for the hippocampus. Among the constraints considered are the limited overall connectivity between the neurons and the distributed processing in a sequence of layered topographically connected maps. Although not all biophysical and modulatory effects from various sources have been incorporated into the present model, the emergent computational function of the hippocampus as a fast storage mechanism with reliable retrieval and pattern completion abilities from partial cues is the main subject of our study. We show that the proposed multiple layer mechanism employing a sparse code and a k-winner-take-all mechanism for the storage and retrieval of binary patterns can be matched to the functional layers of the hippocampus, thereby predicting computational roles for each map and an overall processing principle.