An efficient feature selection method for mobile devices with application to activity recognition

This paper presents a feature selection method for data classification, which combines a model-based variable selection technique and a fast two-stage subset selection algorithm. The relationship between a specified (and complete) set of candidate features and the class label is modelled using a non-linear full regression model which is linear-in-the-parameters. The performance of a sub-model measured by the sum of the squared-errors (SSE) is used to score the informativeness of the subset of features involved in the sub-model. The two-stage subset selection algorithm approaches a solution sub-model with the SSE being locally minimized. The features involved in the solution sub-model are selected as inputs to support vector machines (SVMs) for classification. The memory requirement of this algorithm is independent of the number of training patterns. This property makes this method suitable for applications executed in mobile devices where physical RAM memory is very limited. An application was developed for activity recognition, which implements the proposed feature selection algorithm and an SVM training procedure. Experiments are carried out with the application running on a PDA for human activity recognition using accelerometer data. A comparison with an information gain based feature selection method demonstrates the effectiveness and efficiency of the proposed algorithm.

Practical automated activity recognition using standard smartphones

This paper describes a simple application for mobile devices which automatically recognizes different physical activity. This application could be used to log exercise sessions for the purpose of aiding weight management or improving sporting performance. © 2012 IEEE.

Clustering-Based Ensemble Learning for Activity Recognition in Smart Homes

Application of sensor-based technology within activity monitoring systems is becoming a popular technique within the smart environment paradigm. Nevertheless, the use of such an approach generates complex constructs of data, which subsequently requires the use of intricate activity recognition techniques to automatically infer the underlying activity. This paper explores a cluster-based ensemble method as a new solution for the purposes of activity recognition within smart environments. With this approach activities are modelled as collections of clusters built on different subsets of features. A classification process is performed by assigning a new instance to its closest cluster from each collection. Two different sensor data representations have been investigated, namely numeric and binary. Following the evaluation of the proposed methodology it has been demonstrated that the cluster-based ensemble method can be successfully applied as a viable option for activity recognition. Results following exposure to data collected from a range of activities indicated that the ensemble method had the ability to perform with accuracies of 94.2% and 97.5% for numeric and binary data, respectively. These results outperformed a range of single classifiers considered as benchmarks.

Human behaviour recognition in data-scarce domains

This paper presents the novel theory for performing multi-agent activity recognition without requiring large training corpora. The reduced need for data means that robust probabilistic recognition can be performed within domains where annotated datasets are traditionally unavailable. Complex human activities are composed from sequences of underlying primitive activities. We do not assume that the exact temporal ordering of primitives is necessary, so can represent complex activity using an unordered bag. Our three-tier architecture comprises low-level video tracking, event analysis and high-level inference. High-level inference is performed using a new, cascading extension of the Rao–Blackwellised Particle Filter. Simulated annealing is used to identify pairs of agents involved in multi-agent activity. We validate our framework using the benchmarked PETS 2006 video surveillance dataset and our own sequences, and achieve a mean recognition F-Score of 0.82. Our approach achieves a mean improvement of 17% over a Hidden Markov Model baseline.

A sequential algorithm for sparse support vector classifiers

Support vector machines (SVMs), though accurate, are not preferred in applications requiring high classification speed or when deployed in systems of limited computational resources, due to the large number of support vectors involved in the model. To overcome this problem we have devised a primal SVM method with the following properties: (1) it solves for the SVM representation without the need to invoke the representer theorem, (2) forward and backward selections are combined to approach the final globally optimal solution, and (3) a criterion is introduced for identification of support vectors leading to a much reduced support vector set. In addition to introducing this method the paper analyzes the complexity of the algorithm and presents test results on three public benchmark problems and a human activity recognition application. These applications demonstrate the effectiveness and efficiency of the proposed algorithm.


--------------------------------------------------------------------------------

A Fast Algorithm for Sparse Support Vector Machines for Mobile Computing Applications

This research presents a fast algorithm for projected support vector machines (PSVM) by selecting a basis vector set (BVS) for the kernel-induced feature space, the training points are projected onto the subspace spanned by the selected BVS. A standard linear support vector machine (SVM) is then produced in the subspace with the projected training points. As the dimension of the subspace is determined by the size of the selected basis vector set, the size of the produced SVM expansion can be specified. A two-stage algorithm is derived which selects and refines the basis vector set achieving a locally optimal model. The model expansion coefficients and bias are updated recursively for increase and decrease in the basis set and support vector set. The condition for a point to be classed as outside the current basis vector and selected as a new basis vector is derived and embedded in the recursive procedure. This guarantees the linear independence of the produced basis set. The proposed algorithm is tested and compared with an existing sparse primal SVM (SpSVM) and a standard SVM (LibSVM) on seven public benchmark classification problems. Our new algorithm is designed for use in the application area of human activity recognition using smart devices and embedded sensors where their sometimes limited memory and processing resources must be exploited to the full and the more robust and accurate the classification the more satisfied the user. Experimental results demonstrate the effectiveness and efficiency of the proposed algorithm. This work builds upon a previously published algorithm specifically created for activity recognition within mobile applications for the EU Haptimap project [1]. The algorithms detailed in this paper are more memory and resource efficient making them suitable for use with bigger data sets and more easily trained SVMs.

DNA-dependent Protein Kinase-mediated Phosphorylation of Protein Kinase B Requires a Specific Recognition Sequence in the C-terminal Hydrophobic Motif

DNA-dependent protein kinase (DNA-PK) has been implicated in a variety of nuclear processes including DNA double strand break repair, V(D)J recombination, and transcription. A recent study showed that DNA-PK is responsible for Ser-473 phosphorylation in the hydrophobic motif of protein kinase B (PKB/Akt) in genotoxic-stressed cells, suggesting a novel role for DNA-PK in cell signaling. Here, we report that DNA-PK activity toward PKB peptides is impaired in DNA-PK knock-out mouse embryonic fibroblast cells when compared with wild type. In addition, human glioblastoma cells expressing a mutant form of DNA-PK (M059J) displayed a lower DNA-PK activity when compared with glioblastoma cells expressing wild-type DNA- PK (M059K) when PKB peptide substrates were tested. DNA- PK preferentially phosphorylated PKB on Ser-473 when compared with its known in vitro substrate, p53. A consensus hydrophobic amino acid surrounding the Ser-473 phospho-acceptor site in PKB containing amino acids Phe at position +1 and +4 and Tyr at position -1 are critical for DNA- PK activity. Thus, these data define the specificity of DNA- PK action as a Ser-473 kinase for PKB in DNA repair signaling.

Conserved amino acid residues found in a predicted cytosolic domain of the lipopolysaccharide biosynthetic protein WecA are implicated in the recognition of UDP-N-acetylglucosamine

WecA, an integral membrane protein that belongs to a family of polyisoprenyl phosphate N-acetylhexosamine-1-phosphate transferases, is required for the biosynthesis of O-specific LPS and enterobacterial common antigen in Escherichia coli and other enteric bacteria. WecA functions as an UDP-N-acetylglucosamine (GlcNAc):undecaprenyl-phosphate GlcNAc-1-phosphate transferase. A conserved short sequence motif (His-Ile-His-His; HIHH) and a conserved arginine were identified in WecA at positions 279-282 and 265, respectively. This region is located within a predicted cytosolic segment common to all bacterial homologues of WecA. Both HIHH279-282 and the Arg265 are reminiscent of the HIGH motif (His-Ile-Gly-His) and a nearby upstream lysine, which contribute to the three-dimensional architecture of the nucleotide-binding site among various enzymes displaying nucleotidyltransferase activity. Thus, it was hypothesized that these residues may play a role in the interaction of WecA with UDP-GlcNAc. Replacement of the entire HIHH motif by site-directed mutagenesis produced a protein that, when expressed in the E. coli wecA mutant MV501, did not complement the synthesis of O7 LPS. Membrane extracts containing the mutated protein failed to transfer UDP-GlcNAc into a lipid-rich fraction and to bind the UDP-GlcNAc analogue tunicamycin. Similar results were obtained by individually replacing the first histidine (H279) of the HIHH motif as well as the Arg265 residue. The functional importance of these residues is underscored by the high level of conservation of H279 and Arg265 among bacterial WecA homologues that utilize several different UDP-N-acetylhexosamine substrates.

The activity of a putative polyisoprenol-linked sugar translocase (Wzx) involved in Escherichia coli O antigen assembly is independent of the chemical structure of the O repeat

During O antigen lipopolysaccharide (LPS) synthesis in bacteria, transmembrane migration of undecaprenylpyrophosphate (Und-P-P)-bound O antigen subunits occurs before their polymerization and ligation to the rest of the LPS molecule. Despite the general nature of the translocation process, putative O-antigen translocases display a low level of amino acid sequence similarity. In this work, we investigated whether complete O antigen subunits are required for translocation. We demonstrate that a single sugar, GlcNAc, can be incorporated to LPS of Escherichia coli K-12. This incorporation required the functions of two O antigen synthesis genes, wecA (UDP-GlcNAc:Und-P GlcNAc-1-P transferase) and wzx (O-antigen translocase). Complementation experiments with putative O-antigen translocases from E. coli O7 and Salmonella enterica indicated that translocation of O antigen subunits is independent of the chemical structure of the saccharide moiety. Furthermore, complementation with putative translocases involved in synthesis of exopolysaccharides demonstrated that these proteins could not participate in O antigen assembly. Our data indicate that recognition of a complete Und-P-P-bound O antigen subunit is not required for translocation and suggest a model for O antigen synthesis involving recognition of Und-P-P-linked sugars by a putative complex made of Wzx translocase and other proteins involved in the processing of O antigen.

VEIDase is a principal caspase-like activity involved in plant programmed cell death and essential for embryonic pattern formation

Plant embryogenesis is intimately associated with programmed cell death. The mechanisms of initiation and control of programmed cell death during plant embryo development are not known. Proteolytic activity associated with caspase-like proteins is paramount for control of programmed cell death in animals and yeasts. Caspase family of proteases has unique strong preference for cleavage of the target proteins next to asparagine residue. In this work, we have used synthetic peptide substrates containing caspase recognition sites and corresponding specific inhibitors to analyse the role of caspase-like activity in the regulation of programmed cell death during plant embryogenesis. We demonstrate that VEIDase is a principal caspase-like activity implicated in plant embryogenesis. This activity increases at the early stages of embryo development that coincide with massive cell death during shape remodeling. The VEIDase activity exhibits high sensitivity to pH, ionic strength and Zn2+ concentration. Altogether, biochemical assays show that VEIDase plant caspase-like activity resembles that of both mammalian caspase-6 and yeast metacaspase, YCA1. In vivo, VEIDase activity is localised specifically in the embryonic cells during both the commitment and in the beginning of the execution phase of programmed cell death. Inhibition of VEIDase prevents normal embryo development via blocking the embryo-suspensor differentiation. Our data indicate that the VEIDase activity is an integral part in the control of plant developmental cell death programme, and that this activity is essential for the embryo pattern formation.

Development of Novel Activity-Based Probes for the Detection of Serine Proteases

Cystic Fibrosis (CF) is a genetic disease featuring a chronic cycle of inflammation and infection in the airways of sufferers. Mutations lead to altered ion transport, which in turn causes dehydrated airways and reduced mucociliary clearance which predisposes the patient to infection, resulting in a severe immune response and tissue destruction (1). Airway dehydration is primarily caused by the hyperabsorption of sodium by the epithelial sodium channel (ENaC) (2). ENaC is activated by the action of a number of predominantly trypsin-like Channel Activating Proteases (CAPs) including prostasin, matriptase and furin (3). Additional proteases known to activate ENaC include human airway trypsin (3), plasmin, neutrophil elastase and chymotrypsin (4).

Activity profiling is a valuable technique which involves the use of small inhibitory molecules called Activity-Based Probes (ABPs) which can be used to covalently label the active site of proteases and provide a range of information regarding its structure, catalytic mechanism, location and function within biological systems. The development of novel ABPs for CAPs, would enhance understanding of the role of these proteases in CF airways disease and in particular their role in ENaC activation and airway dehydration. This project investigates the application of a range of novel broad-spectrum ABPs targeting the various subclasses of serine proteases, to include those proteases involved in ENaC activation. Additionally, the application of more selective ABPs in detecting specific serine proteases is investigated.

Compounds were synthesised by Solid-Phase Peptide Synthesis (SPPS) using a standard Fmoc/tBu strategy. Kinetic evaluation of synthesised ABPs against various serine proteases was determined by fluorogenic steady-state enzyme assays. Furthermore, application of ABPs and confirmation of irreversible nature of the compounds was carried out through SDS-PAGE and electroblotting techniques.

Synthesised compounds showed potent irreversible inhibition of serine proteases within their respective targeting class (NAP855 vs Trypsin k3/Ki = 2.60 x 106 M-1 min-1, NFP849 vs Chymotrypsin k3/Ki = 1.28 x 106 M-1 min-1 and NVP800 vs Neutrophil Elastase k3/Ki = 6.41 x 104 M-1 min-1). Furthermore ABPs showed little to no cross-reactivity between classes and so display selectivity between classes. The irreversible nature of compounds was further demonstrated through labelling of proteases, followed by separation and detection via SDS-PAGE and electroblotting techniques. Targeted labelling of active proteases only, was demonstrated by failure of ABPs to detect previously inactivated proteases. Extension of the substrate recognition site within probes resulted in an increased potency and selectivity in the detection of the target proteases. Successful detection of neutrophil elastase from CF sputum samples by NVP800, demonstrated the application of compounds within biological samples and their potential use in identifying further proteases involved in ENaC activation and airway dehydration in CF patients.