Pavel Mautner

Signature verification using ART-2 neural network

The ART neural network models have been developed for the clustering of input vectors and have been commonly used as unsupervised learned classifiers. We describe the use of the ART-2 neural network model for signature verification. The biometric data of all signatures were acquired by a special digital data acquisition pen and fast wavelet transformation was used for feature extraction. The part of authentic signature data was used for training the ART verifier. The architecture of the verifier and achieved results are discussed and ideas for future research are also suggested.

HMM based handwritten text recognition using biometrical data acquisition pen

Development of new text and graphical input devices is considered to be important part of human-computer interaction by many researchers worldwide. The paper presents our experience with the text recognition methods that we have developed for a new designed electronic pen that produces signals corresponding to the movement of the pen on paper. Signals are described by a set of primitives and hidden Markov models are used for word recognition. Results of tests are discussed as well as other possible application areas of our electronic pen.

Auditory ERPs in Children with Developmental Coordination Disorder

The present study aims to investigate and compare the auditory attention performance of children with developmental coordination disorder (DCD) and normally developing children (NDC) using cognitive evoked potentials (ERPs) in passive conditions. ERPs data showed that children with DCD have less ability to detect small physical differences between acoustic stimuli (no MMN response in DCD children) and have a reduced attentional engagement and stimulus evaluation of salient stimuli (a reduction of P3 amplitude in DCD children). The results of our study suggest that children with DCD do not only suffer from a visuospatial attention deficit as previous studies reported but also have auditory attention deficit.

Off-line analysis of the P300 event-related potential using discrete wavelet transform

The P300 component is an event-related potential associated with the process of decision making. Its accurate detection in electroencephalographic signal is essential for P300-based brain-computer interfaces. This paper proposes a new method for the P300 detection based on discrete wavelet transform. Its benefits were tested with five healthy subjects using voting classification algorithm. The accuracy of 74.4% was achieved for single trials and over 90% accuracy for averaging.

Using the Windowed means paradigm for single trial P300 detection

The Windowed means paradigm is a method for slow-changing cortical potentials feature extraction, most importantly in reaction to events. It has been successfully used for various brain-computer interfaces. The objective of this paper was to evaluate if this paradigm is also appropriate for P300 brain-computer interfaces. The modified method was tested on five healthy subjects. The optimal selection of parameters was discussed. The Windowed means paradigm was successful for the P300 detection on the testing data-set.

The event-related potential data processing using ART 2 network

The event-related potentials (ERPs) obtained by stimulation are much weaker than the continuous electro-encephalographic (EEG) signal. Therefore, the correct signal analysis is vital to detect the stimulation-driven signal components. This paper proposes the combination of matching pursuit for feature extraction and ART 2 neural network for clustering. Then, clusters are filtered and interpreted according to their statistical properties as ERP components or noise. The suggested method can be used to filter the EEG/ERP signal. Furthermore, its results lead to a method that improves averaging when compared to traditional approaches.

PROCESSING AND CATEGORIZATION OF CZECH WRITTEN DOCUMENTS USING NEURAL NETWORKS

The Kohonen Self-organizing Feature Map (SOM) has been developed for clustering input vectors and for projection of continuous high-dimensional signal to discrete low-dimensional space. The application area, where the map can be also used, is the processing of text documents. Within the project WEBSOM, some methods based on SOM have been developed. These methods are suitable either for text documents information retrieval or for organization of large document collections. All methods have been tested on collections of English and Finnish written documents. This article deals with the application of WEBSOM methods to Czech written documents collections. The basic principles of WEBSOM methods, transformation of text information into the real components feature vector and results of documents classification are described. The Carpenter-Grossberg ART-2 neural network, usually used for adaptive vector clustering, was also tested as a document categorization tool. The results achieved by using this network are also presented.

WEBSOM Method - Word Categories in Czech Written Documents

We applied well-known WEBSOM method (based on two layer architecture) to categorization of Czech written documents. Our research was focused on the syntactic and semantic relationship within word categories of word category map (WCM). The document classification system was tested on a subset of 100 documents (manual work was necessary) from the corpus of Czech News Agency documents. The result confirmed that WEBSOM method could be hardly evaluated because humans have problems with natural language semantics and determination of semantic domains from word categories.

Cognitive impairment measured by event-related potentials during early and late postoperative period following intravenous or inhalation anaesthesia

OBJECTIVE This study investigated modification in cognitive function following inhalation (IA) and total intravenous (TIVA) anaesthesia measured using auditory ERPs (Event Related Potentials). METHODS Auditory ERPs examination with N1, P3a and P3b component registration was carried out one day before surgery (D-1) and on the first (D+1), sixth (D+6) and 42nd (D+42) days after surgery. Results were compared between two anaesthetic groups. RESULTS On D+1, N1 latency was increased in the IA group. A significant reduction was observed in amplitude of the P3a component on D+6, which persisted up to D+42 for both IA and TIVA groups. A reduction in the amplitude of P3b on D+1 with normalization by D+6 was found in both groups as well. CONCLUSIONS Intravenous and inhalation anaesthesia lead to similar changes in cognitive function as determined by ERPs, both during the early and late postoperative periods. It cannot be clearly confirmed whether the observed effects are due to anaesthesia or other unmonitored perioperative factors. SIGNIFICANCE Post anaesthetic changes represent a subclinical impairment; nevertheless, they represent a potential risk for subsequent development of cognitive difficulties.

T113. The changes in motor cortex localisation and organisation in motor eloquent tumorous brain lesions detected by navigated transcranial magnetic stimulation

Introduction Navigated transcranial magnetic stimulation (nTMS) is non invasive method to map the motor cortex including primary motor cortex (PrG - precentral gyrus) and premotor areas (PMa). This study aimed to investigate whether tumorous brain lesion induce a change in motor cortex localization or organisation investigated by nTMS. Methods We enrolled 10 patients with intraaxial motor tumor (gliomas). All patients underwent preoperative navigational MRI folowed by nTMS. Both lesional and non lesional hemispheres were stimulated. MEPs were recorded by EMG. The measured muscle was APB. MEPs latency for each positive stimulation point was measured. The surface of positive stimulation areas on the cortex were calculated for both hemispheres and results were compared. Results The positive MEPs responses were registered from PrG as well as from PMa with different latencies. There were mosaic distribution of short and long latencies of MEPs responses without dependence on the location in the PrG or PMa in both hemispheres. The positive motor area distributions were significantly larger from lesional than for non lesional hemisphere for both PrG and PMa areas (PrG lesional vs. non lesional surface = 947 mm2 vs. 393 mm2, PMa lesional vs. non lesional surface = 620 mm2 vs. 545 mm2). Conclusion The intraaxial motor eloquent tumors induce changes in motor cortex. The motor areas spread widely in the anterior-posterior direction in lesional hemisphere. The localisation short latencies found in PMa suggesting for detection of primary motor areas outside the PrG. This study was supported by the Charles University Research Fund Progress Q 39.