Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Steady-state visually evoked potentials: focus on essential paradigms and future perspectives.

Brain-computer interface (BCI) systems based on the steady-state visual evoked potential (SSVEP) provide higher information throughput and require shorter training than BCI systems using other brain signals. To elicit an SSVEP, a repetitive visual stimulus (RVS) has to be presented to the user. The RVS can be rendered on a computer screen by alternating graphical patterns, or with external light sources able to emit modulated light. The properties of an RVS (e.g., frequency, color) depend on the rendering device and influence the SSVEP characteristics. This affects the BCI information throughput and the levels of user safety and comfort. Literature on SSVEP-based BCIs does not generally provide reasons for the selection of the used rendering devices or RVS properties. In this paper, we review the literature on SSVEP-based BCIs and comprehensively report on the different RVS choices in terms of rendering devices, properties, and their potential influence on BCI performance, user safety and comfort.

/pdf/a-survey-of-stimulation-methods-used-in-ssvep-based-bcis-38omyw8mqs.pdf

A survey of stimulation methods used in SSVEP-based BCIs

In this paper, novel methods for detecting steady-state visual evoked potentials using multiple electroencephalogram (EEG) signals are presented. The methods are tailored for brain-computer interfacing, where fast and accurate detection is of vital importance for achieving high information transfer rates. High detection accuracy using short time segments is obtained by finding combinations of electrode signals that cancel strong interference signals in the EEG data. Data from a test group consisting of 10 subjects are used to evaluate the new methods and to compare them to standard techniques. Using 1-s signal segments, six different visual stimulation frequencies could be discriminated with an average classification accuracy of 84%. An additional advantage of the presented methodology is that it is fully online, i.e., no calibration data for noise estimation, feature extraction, or electrode selection is needed

Multiple Channel Detection of Steady-State Visual Evoked Potentials for Brain-Computer Interfaces

A textbook of physiology

A popular speech synthesis method is the HMM based speech synthesis method. Given the phone set and question set for a language, HMM based synthesis systems are built. Although robotic in quality the systems are intelligible. In this paper, we propose a common framework for Indian languages with a common phone set and a common question set. Owing to this architecture it is possible to borrow independent monophone models across languages. Degradation MOS and word error rate scores are comparable to systems built in the conventional language-specific manner, indicating that system building can be made language-independent without much degradation in the quality of synthesised speech.

An approach to building language-independent text-to-speech synthesis for Indian languages

In this study, a high performance electrooculogram (EOG) based synchronous and asynchronous visual keyboard system is designed with large number of targets. The proposed system overcomes the limitations in the conventional EEG based visual keyboard system which has to compromise with either accuracy or typing speed of the system. In this study, we first proposed a synchronous visual keyboard system with 36 targets which includes alphabets, numbers and space. Later, we proposed an asynchronous visual keyboard system where the user can have control on the system. Ten subjects were taken for evaluating the performance of the proposed keyboard systems. A cue guided online experimental procedure was conducted on all the subjects in synchronous and asynchronous case. The average classification accuracy of 94.2% and 98.79% were obtained in synchronous and asynchronous cases respectively. Four subjects show 100% accuracy in the asynchronous case with average information transfer rate (ITR) of 66.71 bits/minute. This result shows that the developed system outperforms conventional keyboard system.

A Novel EOG based Synchronous and Asynchronous Visual Keyboard System

Coronary artery disease producing ischemic cardiomyopathy is the most frequent cause of left ventricular systolic dysfunction. Non-ischemic cardiomyopathies can also produce systolic dysfunction; they may be inherited as genetic disorders or occur sporadically. These coronary injuries have repercussions on the left ventricle producing changes on wall contractility, the shape of the cavity and also changes on ventricular function. This study is focused on the 2D echocardiograms of the left ventricle. Apical two chamber and four chamber view recordings were performed on normal and systolic dysfunction subjects. Individual frames were extracted for at least five cardiac cycles. After preprocessing these images, segmentation of the left ventricle was performed by Fuzzy systems. Then the volumes were measured by single and biplane methods along with the perimeter, short axis length and long axis length in each frame, from which the two indices Sphericity Index (SI) and Normalized Eccentricity Index (NEI) was determined. It was found that the diastolic phase is short in the case of systolic dysfunction, and its volume variation is not uniform as in the normal case. Also, in the case of systolic dysfunction, the span of either the long or short axis length variation is less than 0.5 cm. This depicts that akinesis is in the corresponding direction; the value of SI is less than 2 for systolic dysfunction. A sharp peak is seen at each systole point in the NEI plot and also its variation is smooth in subjects having LVEF > 45%, which is not the case for dysfunction.

Detection of left ventricle systolic dysfunction from shape deformity.

The recently proposed compressed sensing theory equips us with methods to recover exactly or approximately, high
resolution images from very few encoded measurements of the scene. The traditional ill-posed problem of MRI image
recovery from heavily under-sampled κ-space data can be thus solved using CS theory. Differing from the soft
thresholding methods that have been used earlier in the case of CS MRI, we suggest a simple iterative hard thresholding
algorithm which efficiently recovers diagnostic quality MRI images from highly incomplete κ-space measurements. The
new multi-scale redundant systems, curvelets and contourlets having high directionality and anisotropy, and thus best
suited for curved-edge representation are used in this iterative hard thresholding framework for CS MRI reconstruction
and their performance is compared. The κ-space under-sampling schemes such as the variable density sampling and the
more conventional radial sampling are experimented at the same sampling rate and the effect of encoding scheme on
iterative hard thresholding compressed sensing reconstruction is studied.

M. Ramasubba Reddy

Papers

An approach to building language-independent text-to-speech synthesis for Indian languages

A Novel EOG based Synchronous and Asynchronous Visual Keyboard System

Detection of left ventricle systolic dysfunction from shape deformity.

An iterative hard thresholding algorithm for CS MRI

Constructing an exactly periodic subspace for enhancing SSVEP based BCI