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

Data Mining - Concepts and Techniques.

Compressive Sensing (CS) is a new sensing modality, which compresses the signal being acquired at the time of sensing. Signals can have sparse or compressible representation either in original domain or in some transform domain. Relying on the sparsity of the signals, CS allows us to sample the signal at a rate much below the Nyquist sampling rate. Also, the varied reconstruction algorithms of CS can faithfully reconstruct the original signal back from fewer compressive measurements. This fact has stimulated research interest toward the use of CS in several fields, such as magnetic resonance imaging, high-speed video acquisition, and ultrawideband communication. This paper reviews the basic theoretical concepts underlying CS. To bridge the gap between theory and practicality of CS, different CS acquisition strategies and reconstruction approaches are elaborated systematically in this paper. The major application areas where CS is currently being used are reviewed here. This paper also highlights some of the challenges and research directions in this field.

A Systematic Review of Compressive Sensing: Concepts, Implementations and Applications

Most image encryption algorithms based on low-dimensional chaos systems bear security risks and suffer encryption data expansion when adopting nonlinear transformation directly. To overcome these weaknesses and reduce the possible transmission burden, an efficient image compression–encryption scheme based on hyper-chaotic system and 2D compressive sensing is proposed. The original image is measured by the measurement matrices in two directions to achieve compression and encryption simultaneously, and then the resulting image is re-encrypted by the cycle shift operation controlled by a hyper-chaotic system. Cycle shift operation can change the values of the pixels efficiently. The proposed cryptosystem decreases the volume of data to be transmitted and simplifies the keys distribution simultaneously as a nonlinear encryption system. Simulation results verify the validity and the reliability of the proposed algorithm with acceptable compression and security performance.

Image compression–encryption scheme based on hyper-chaotic system and 2D compressive sensing

The existing ways to encrypt images based on compressive sensing usually treat the whole measurement matrix as the key, which renders the key too large to distribute and memorize or store. To solve this problem, a new image compression–encryption hybrid algorithm is proposed to realize compression and encryption simultaneously, where the key is easily distributed, stored or memorized. The input image is divided into 4 blocks to compress and encrypt, then the pixels of the two adjacent blocks are exchanged randomly by random matrices. The measurement matrices in compressive sensing are constructed by utilizing the circulant matrices and controlling the original row vectors of the circulant matrices with logistic map. And the random matrices used in random pixel exchanging are bound with the measurement matrices. Simulation results verify the effectiveness, security of the proposed algorithm and the acceptable compression performance.

Novel image compression–encryption hybrid algorithm based on key-controlled measurement matrix in compressive sensing

Successive Projections onto INcoherent manifolds (SPIN) recovers signals from random measurements, provided that signal manifolds are incoherent and the measurement operator satisfies Restricted Isometric Property. This paper proposes a dictionary based method for recovering non-parametric signal mixtures from a few random projections with individual components on incoherent manifolds. The proposed method is an extended version of SPIN and can be applied to separate signal mixtures theoretically without any limit on the number of components.

A dictionary based approach for non-parametric SPIN and application to image mixture separation

An infrared thermal imaging system can screen large groups of people quickly, non-intrusively and with a high level of confidence in spotting febrile subjects. A thermal imaging system must identify individuals with a body core temperature above 38◦C to effectively do this. Conventionally, this is per-formed using thermal imaging cameras in the Long Wave Infra Red (LWIR) region. But the issue with most of these systems is that it fails to read the temperature at the inner eye canthi, which is the ideal location on the face for fever detection. This is owing to the lack of distinguishing features that are available on a thermal image. Hence, in this paper, an attempt is made to read the temperature at the inner eye canthus by first detecting the eye region by running the Viola - Jones Haar Cascade algorithm over the visible images, in a popular synchronized visible and thermal face dataset. Subsequently, this identified region of interest in the visible image is registered against its corresponding thermal image using the ORB algorithm to locate the eye canthus area on the thermal image. Experimental results analysed in this paper proves that a read out of the temperature at this pixel location can be a sufficiently accurate estimate of the core body temperature of the subject being screened.

Cross Spectral Image Registration for Fever Detection

Human Activity Recognition (HAR) is one of the most active research areas in fields of computer vision and pattern analysis. Most of the existing HAR algorithms are devised in supervised manner by excluding the temporal aspects of skeletal data which is a key parameter in HAR. Motivated by this, we have designed and developed an efficient subspace clustering algorithm for HAR by explicitly considering the time series aspects of human activity data. Designing this algorithm in an unsupervised manner is another challenge that we are dealing with. The work involves design of an efficient covariance descriptor for encoding the skeletal data. Later a subspace clustering algorithm called temporal subspace clustering (TSC) algorithm is designed by exploiting the principles of Laplacian regularization and dictionary learning. Experimental analysis shows that the proposed method outperforms the state-of-the-art methods employed for HAR.

Human Activity Recognition from Skeletal Data using Covariance Descriptor and Temporal Subspace Clustering

The paper discusses a method for improving the SNR and hence sensitivity of Frequency Modulated Continuous Wave (FM-CW) radar. In FM-CW radar, a beat signal is generated internally, with its frequency proportional to the altitude. Beat frequency measurement results in error if the input has low Signal to Noise Ratio (SNR). SNR and hence sensitivity of the receiver can be improved using digital tracking filter. This can be achieved using a set of pre-designed band pass filters (BPF) or a suitable adaptive filter. In practical application, the statistical characteristics of signal and noise are usually unknown. The design of fixed filter is based on prior knowledge of both the signal and noise. Adaptive filters, on the other hand, have the ability to adjust their own parameters automatically and have immense potential in radar applications. The adaptive line enhancer (ALE) is applied before the FFT analysis to improve the detection performance. Performance of ALE using LMS, RLS algorithms and a set of block processing algorithms BLMS, BAP are compared. Adaptive filters with different algorithms were simulated in Visual DSP++ before implementing in hardware. The results show that ALE algorithms can be used as tracking filter for SNR enhancement. Processing speed of ALE algorithms were compared using Analog Devices Signal Processor 21060. Block LMS give better performance compared to other algorithms. Tracking filter using ALE BLMS was implemented in signal processing board. Results show that tracking filter using ALE improves the sensitivity of the system.

Digital tracking filter for FM-CW radar

A multi-view image dataset is highly correlated and redundant. In this paper, we propose a multi-view image compression technique which exploits the inter-frame correlation in a dataset. A frame is divided into patches and sparse coding is applied to each patch, utilising an over-complete dictionary derived from a highly correlated region in the preceding frame and dictionary atoms are built as vector representation overlapping patches of selected region. The degree of sparsity of each patch can be controlled to achieve specified rate-distortion performance. The proposed technique does not require storage or transmission of dictionary atoms and all the frames in the dataset can be compressed by keeping preceding frame as reference. Performance of the proposed compression scheme is compared with JPEG2000 & Depth Layer based techniques and results reveal that proposed scheme outperforms it.

Sudhish N. George

Papers

A dictionary based approach for non-parametric SPIN and application to image mixture separation

Cross Spectral Image Registration for Fever Detection

Human Activity Recognition from Skeletal Data using Covariance Descriptor and Temporal Subspace Clustering

Digital tracking filter for FM-CW radar

Local patch dictionary based approach for multi-view image compression