Mihaela Costin

An EEG Coherence Based Method Used for Mental Tasks Classification

In this paper a new coherence based method to extract the appropriate EEG features for a five mental tasks classification problem is proposed. The new introduced method has the advantage of using an adaptive new technique that models the EEG signal using the frequency information obtained by employing the coherence function to each EEG recording channel. The adaptive attribute of the technique is due to both, to the amplitude and, respective, to the phase adaptive processes used to model the EEG signal. Another specificity of the new modeling technique is given by the fact of exploiting the nonlinear dynamics of the brain system; this is reflected in the particular spectral mixing of the fundamental spectral components obtained first, by using the coherence function. Finally, to conclude the obtained results in comparison with the results reported in the literature, by using this new approach the classification rate was noticeably improved.

Using neural networks and LPCC to improve speech recognition

Linear Predictive Coding (LPC), powerful speech analysis technique, is very useful for encoding speech at a low bit rate and provides extremely accurate estimates of speech parameters - based on the assumption that speech signal is produced by a buzzer at the end of the tube (the glottis produces the buzz, characterized by its intensity and frequency, and the vocal tract forms the tube, characterized by resonance frequencies (formants) according to Calliope(1989), is very efficient for the vocalic areas. The model is less efficient for transient, unvowel or not stationary regions according to R. Lawrence and B. Hwang Juang (1993). A Radial Basis Function network is able to recognize in a satisfying percent a set of phonemes pronounced by different speakers, using LPC sets as input.

Image Categorization Based on Computationally Economic LAB Colour Features

An easy to compute and small colour feature vector is introduced in this paper, as a tool to be used in the process of retrieval or classification of similarly coloured digital images from very large databases. A particular set of “ab” planes from the LAB colour system is used, along with a specific configuration of colour regions within them. The colour feature vector is low dimensional (only 96 components), computationally economic and performs very well on a carefully selected database of rose images, publicly available.

Optic Disc Localization in Retinal Images

The paper proposes an optic disc localisation method in color retinal images. It is a first step of a retinal image analysis project which will be completed later with other tasks as fovea detection and measurement of retinal vessels. The final goal is to detect in early stages signs of ophthalmic pathologies as diabetic retinopathy or glaucoma, by successive analysis of ophthalmoscopy images. The proposed method first detects in the green component of RGB image the optic disc area and then on the segmented area extracts the optic disc edges and obtains a circular optic disc boundary approximation by a Hough transform.

3D Breast Shape Reconstruction for a Non-Invasive Early Cancer Diagnosis System

Early breast cancer diagnosis becomes a must in nowadays society, related to the fact that this osteophyl neoplasia is the most frequent cancer in the world. The environment, solar radiations, weather changes, the increasing number of endocrine modifications, iatrogen agression, stress, or even fashion, have undoubtful effects on womens breasts health, too. Long term negative influence may appear due to invasive or even minim invasive cancer investigation techniques, so, alternative methods are continuously researched. As a stage in the implementation of a non-invasive microwaves technique for breast cancer diagnosis under research, meant to early uncover the developing malignant processes, we designed a 3D reconstruction process that we are extensively presenting in this paper. This reconstruction is a compulsory step previous to the tumor volumetric shape reconstruction, in order to be well positioned for further breast surgery, where our results could find a field of application. I.

A novel iris clustering approach using LAB color features

Interesting results of color clustering for the iris images in the UBIRISv1 database are presented. The iris colors are characterized by feature vectors with 80 components corresponding to histogram bins computed in the CIELAB color space. The feature extraction is applied to the first session eye images after undergoing an iris segmentation process. An agglomerative hierarchical algorithm is used to organize 1.205 segmented iris images in 8 clusters based on their color content.

New attempts in sound diarization

The paper discuses a new hybrid method in sound diarization (the process of segmenting an audio file into chunks that represent unique sources and clustering the obtained segments into groups that represent the same item). The most recent results are focusing mainly on the identification of voices during the telephonic recordings. In the hybrid method proposed here, a clustering is applied first, using an agglomerative approach regarding the construction of speaker models. Subsequently, when consistent amounts of data are gathered, special models are built using speaker factors. This idea gives good performance over the classical approach as the low-level clustering Bayesian Information Criterion scheme has poor performance on complex models, where speaker factors have very good precision. Speaker diarization improves speaker verification for multi-speaker audio (summed channel telephone data, single microphone interview data), is very important for speech recognition, and improves readability of an automatic transcription by structuring the audio stream into speaker turns and in some cases by providing the identity of the speakers. Sound diarization offers information which can be of interest for the multimedia documents indexing, in human-computer interaction, robotics, security systems, etc.

Improving cochlear implant performances by MFCC technique

Cochlear Implant (CI) is a device meant to recover hearing abilities for patients suffering of total bilateral cophosis. In this study we try to identify phonemes that usually have a low CI recognition rate, using mel-frequency cepstral coefficients (MFCC) technique. In order to analyze them we had to structure a special signal database: we registered phonemes passed by the CI testing device. By a technique of accentuating certain frequency bands - depending on the recognized phoneme - we intend to improve CI performances. Clustering is realized by the help of a usual two-layer MLP neural network. In parallel, we extract, using the same step as in the cochlear implant device technique already implemented, a new set of MFCC coefficients from speech, and compare them. Using fuzzy functions in computing energy on frequency bands results are slightly better according to M. Costin et al.(2002).

Data Flow Chart in a Non-Invasive Breast Cancer Diagnosis System

Breast cancer continues to be the most common of nonpreventable cancer diagnosed among women. We propose a non-invasive method of abnormal area detection based on microwaves emission detection, in order to uncover the incipient malignant processes. Underlying the actual situation, statistics and methods, our paper comments the necessary stages implemented in order to distinguish and compare the body normal and abnormal microwave radiations. In the same time, we highlight some problems, punctually submitting a fuzzy solution. This paper presents the starting point and some important, necessary steps in our research, which we are further continuing.

Improving Noninvasive Monitoring in Medical Care

Early detection of non-uniformities and abnormalities in human body behavior could save lives and might diminish health care assurance expenses. The range of monitored parameters is wide, from respiratory and cardio parameters survey to early cancer detection. Due to invasive methods important drawbacks, researchers are looking for new non-invasive methods. The is presenting new approaches, combining different physical methods, compared to the existent monitoring techniques, usually used in medical care. Our approach concerns studies of using microwave Doppler effect for the contactless detection of the cardiac and respiratory activity, for the continuous respiratory and cardiac monitoring. We are also describing a microwave body emission radiometry method in abnormal tissue increased activity detection.