Asha Vijayan

Audio Steganography Using Modified LSB and PVD

In Audio Steganography we find a way so that an audio file can be used as a host media to hide textual message without affecting the file structure and content of the audio file. In this system a novel high bit rate LSB audio data hiding and another method known as Pixel value differencing is proposed. This scheme reduces embedding distortion of the host audio. The hidden bits are embedded into the higher LSB layers resulting in increased robustness against noise addition. To avoid major differences from the cover audio and the embedded audio this algorithm helps in modifying the rest of the bits. To enlarge the capacity of the hidden secret information and to provide an imperceptible stego-audio for human perception, a pixel-value differencing (PVD) is used for embedding. The difference value of audio samples is replaced by a new value to embed the value of a sub-stream of the secret message. The method is designed in such a way that the modification is never out of the range interval. This method provides an easy way to produce a more imperceptible result than those yielded by simple least-significant-bit replacement methods. The SNR value is good for LSB scheme and the capacity is high for PVD scheme.

Computationally EfficientBio-realistic Reconstructions of Cerebellar Neuron Spiking Patterns

Simple spiking models have been known to replicate detailed mathematical models firing properties with reliable accuracy in spike timing. We modified the adaptive exponential integrate and fire mathematical model to reconstruct different cerebellar neuronal firing patterns. We were able to reconstruct the firing dynamics of various types of cerebellar neurons and validated with previously published experimental studies. To model the neurons, we exploited particle swarm optimization to fit the parameters. The study showcases the match of electroresponsiveness of the neuronal models to data from biological neurons. Results suggest that models are close reconstructions of the biological data since frequency and spike-timing closely matched known values and were similar to those in previously published detailed computationally intensive biophysical models. Such spiking models have a number of applications including design of large-scale circuit models in order to understand physiological dysfunction and for various computational advantages.

Wireless intrusion detection based on different clustering approaches

Wireless security is becoming an important area of product research and development. Wireless Intrusion detection Systems are commonly used in WLAN network for detecting wireless attacks. Classifiers are commonly used as detectors in these systems. Finding an efficient classifier as well selecting best set of features becomes very important for implementing these intrusion detection systems. In this paper, we are finding optimital set of features from collected WLAN data using a Ranking Algorithm technique. Then with the aid of different data mining techniques such as K-Means, self organizing map and decision tree, these features are analyzed and the performance comparison is carried out.

Neural Control using EEG as a BCI Technique for Low Cost Prosthetic Arms

There have been significant advancements in brain computer interface (BCI) techniques using EEG-like methods. EEG can serve as non-invasive BMI technique, to control devices like wheelchairs, cursors and robotic arm. In this paper, we discuss the use of EEG recordings to control low-cost robotic arms by extracting motor task patterns and indicate where such control algorithms may show promise towards the humanitarian challenge. Studies have shown robotic arm movement solutions using kinematics and machine learning methods. With iterative processes for trajectory making, EEG signals have been known to be used to control robotic arms. The paper also showcases a case-study developed towards this challenge in order to test such algorithmic approaches. Non-traditional approaches using neuro-inspired processing techniques without implicit kinematics have also shown potential applications. Use of EEG to resolve temporal information may, indeed, help understand movement coordination in robotic arm.

Classification of robotic arm movement using SVM and Naïve Bayes classifiers

Target-oriented approaches have been commonly used in robotics. In 3D space, movement of a robotic arm depends on the target position which can either follow a forward or inverse kinematics approach to reach the target. Predicting the movement of a robotic arm requires prior learning through methods such as transformation matrices or other machine learning techniques. In this paper, we built an online robotic arm to extract movement datasets and have used machine learning algorithms to predict robotic arm articulation. For efficient training, small training datasets were used for learning purpose. Classification is used as a scheme to replace prediction-correction approach and to test whether the method can function as a replacement of usual forward kinematics schemes or predictor-corrector methods in directing a remotely controlled robotic articulator. This study reports significant classification accuracy and efficiency on real and synthetic datasets generated by the device. The study also suggests linear SVM and Naïve Bayes algorithms as alternatives for computational intensive learning schemes while predicting articulator movement in laboratory environments.

Computational Neuroscience of Timing, Plasticity and Function in Cerebellum Microcircuits

Cerebellum has been known to show homogeneity in circuit organization and hence the “modules” or various circuits in the cerebellum are attributed to the diversity of functions such as timing, pattern recognition, movement planning and dysfunctions such as ataxia related to the cerebellum. Ataxia-like conditions, induced by intrinsic excitability changes, disable spiking or bursts and thereby limit the quanta of downstream information. Understanding timing, plasticity and functional roles of cerebellum involve large-scale and microcircuit reconstructions validating molecular mechanisms in population activity. Using mathematical modelling, we attempted to reconstruct information transmission at the granular layer of the cerebellum, a circuit whose role in dysfunctions remain yet to be fully explored. We have employed spiking models to reconstruct timing roles and detailed biophysical models for extracellular activity and local field population response. The roles of inhibition, induced plasticity and their implications in information transmission were evaluated. Modulatory roles of Golgi inhibition and pattern abstraction via optimal storage were estimated. An abstraction of the granular and Purkinje layer circuit for neurorobotic roles such as pattern recognition and spike encoding via two new methods was developed. Simulations suggest plasticity at cerebellar relays may be an important element of tremendous storage capacity reliable in the learning of coordination of actions, sensorimotor or cognitive, in which the cerebellum participates.

Low-cost robotic articulator as an online education tool: Design, deployment and usage

Humanitarian challenges in developing nations such as low cost prosthesis for the physically challenged, have also led to substantial progress in robotics. In this paper, we implemented and deployed a low-cost remotely controlled robotic articulator, as an education tool for university students and teachers. This tool is freely available online and is being employed to generate robotic datasets for novel algorithms. Using a server-client methodology and a browser-based user interface, the online lab allows learners to access and perform basic kinematics experiments and study robotic articulation. These experiments were developed for allowing students to enhance laboratory skills in robotics and improve practical experience without concerns for equipment access restrictions or cost.

Spike encoding for pattern recognition: Comparing cerebellum granular layer encoding and BSA algorithms

Spiking neural encoding models allow classification of real world tasks to suit for brain-machine interfaces in addition to serving as internal models. We developed a new spike encoding model inspired from cerebellum granular layer and tested different classification techniques like SVM, Naïve Bayes, MLP for training spiking neural networks to perform pattern recognition tasks on encoded datasets. As a precursor to spiking network-based pattern recognition, in this study, real world datasets were encoded into spike trains. The objective of this study was to encode information from datasets into spiking neuron patterns that were relevant for spiking neural networks and for conventional machine learning algorithms. In this initial study, we present a new approach similar to cerebellum granular layer encoding and compared it with BSA encoding techniques. We have also compared the efficiency of the encoded dataset with different datasets and with standard machine learning algorithms.

Modeling pattern abstraction in cerebellum and estimation of optimal storage capacity

Precise fine-tuning of motor movements has been known to be a vital function of cerebellum, which is critical for maintaining posture and balance. Purkinje cell (PC) plays a prominent role in this fine-tuning through association of inputs and output alongside learning through error correction. Several classical studies showed that PC follows perceptron like behavior, which can be used to develop cerebellum like neural circuits to address the association and learning. With respect to the input, the PC learns the motor movement through update of synaptic weights. In order to understand how cerebellar circuits associate spiking information during learning, we developed a spiking neural network using adaptive exponential integrate and fire neuron model (AdEx) based on cerebellar molecular layer perceptron-like architecture and estimated the maximal storage capacity at parallel fiber-PC synapse. In this study, we explored information storage in cerebellar microcircuits using this abstraction. Our simulations suggest that perceptron mimicking PC behavior was capable of learning the output through modification via finite precision algorithm. The study evaluates the pattern processing in cerebellar Purkinje neurons via a mathematical model estimating the storage capacity based on input patterns and indicates the role of sparse encoding of granular layer neurons in such circuits.

Comparing robotic control using a spiking model of cerebellar network and a gain adapting forward-inverse model

Internal models inspired from the functioning of cerebellum are being increasingly used to predict and control movements of anthropomorphic manipulators. A major function of cerebellum is to fine tune the body movements with precision and are comparative to capabilities of artificial neural network. Several studies have focused on encoding the real-world information to neuronal responses but temporal information was not given due importance. Spiking neural network accounts to conversion of temporal information into the adaptive learning process. In this study, cerebellum like network was reconstructed which encodes spatial information to kinematic parameters, self-optimized by learning patterns as seen in rat cerebellum. Learning rules were incorporated into our model. Performance of the model was compared to an optimal control model and have evaluated the role of bioinspired models in representing inverse kinematics through applications to a low cost robotic arm developed at the lab. Artificial neural network of Kawato was used to compare with our existing model because of their similarity to biological circuit as seen in a real brain. Kawatos paired forward inverse model has used to train for fast movement based tasks which resembles human based motor tasks. Result suggest kinematics of a 6 DOF robotic arm was internally represented and this may have potential application in neuroprosthesis.