Due to the increase in technology and research over the past few decades, music had become increasingly available to the public, but with a vast selection available, it becomes challenging to choose the songs to listen too. From research done on music recommendation systems (MRS), there are three main methods to recommend songs; context based, content based and collaborative filtering. A hybrid combination of the three methods has the potential to improve music recommendation; however, it has not been fully explored. In this paper, a hybrid music recommendation system, using emotion as the context and musical data as content is proposed. To achieve this, the outputs of a convolution neural network (CNN) and a weight extraction method are combined. The CNN extracts user emotion from a favorite playlist and extracts audio features from the songs and metadata. The output of the user emotion and audio features is combined, and a collaborative filtering method is used to select the best song for recommendation. For performance, proposed recommendation system is compared with content similarity music recommendation system (CSMRS) as well as other personalized music recommendation systems.

Significant improvements can be made by integrating technology in training as shown in various athletics such as football, soccer, and tennis. In track and field, specifically throwing events, training is still done in the traditional way using just a coach’s instruction. With no method to measure specific data, an athlete cannot always perform to the best of their ability. A system was developed during this project to measure and improve the training of a throwing athlete using motion capture and virtual reality. This involves capturing the rotational and movement data of an athlete, sending it in a viable format to the cloud using quaternions. Creating an armature that represents the athlete in a virtual environment, receiving the movement data from the cloud and applying it to the armature. From this, the movement of the athlete can be studied using a virtual reality headset and using a chart that shows values such as velocity, speed, and acceleration. The results of the project showed a noticeable lag in real time caption, and a drift after a period. Overall, a video of the motion capture was useful in analyzing an athlete movement.

The goal of this research is to estimate the degree to which a person appreciates a given piece of music, predict and recommend the next music piece/song. Listening to music is subjective and dynamic. Such profiles may be helpful in applications like digital forensics. For example, prisoners, people with depression or other health issues. This research work involves designing a deep learning algorithm to extract attributes of music, such as tempo, and pitch, etc. A set of such attributes is identified using attribute classification stage, to define the liking index (L-i) based on the user responses, subsequently appearing as inputs to the neural network. The learning capability of the neural networks is employed to build the music profiles in response to the inputs. The attribute classification stage involves clustering using the K-means algorithm for the created music profiles. Synthetic data (user-selected set of music pieces) was created initially for testing the proof of concept. Preliminary results show that for K = 5, the classifier results show minimum outliers with well-defined clusters. Real data and steps to validate the proposed algorithm are the next steps to be carried out. After validation, the model will be tested with user data to confirm the design and its effectiveness.

This work presents our research aimed to develop a driver safety assistant system. The idea is to use in-vehicle camera with vision sensor to detect emotional distress level of the driver while driving. An algorithm to identify facial expression of the driver is developed using Python programming language. In addition, a prototype of facial expression detection along with a car parking assist system is developed by using an Arduino Uno ATMEL ATMEGA328 microcontroller interfaced with a webcam and a motor to demonstrate the concept. The camera mounted on the dashboard continuously monitors the driver’s face and captures the facial expressions. The facial expressions so captured help assess the driver’s (particularly, the truck driver) situation and identify it in terms of severe pain, headache, cardiac arrest, etc. Once the system identifies the situation, controller then assists in driving the car to the curb and bringing it to a complete stop. The facial expression identification algorithm uses the sensors (like speed, steering etc.) to detect the abnormality from the facial expression and subsequently alert the driver for 30 s. The system continuously checks the driver’s profile. If the driver is driving while continuously in pain for another 30 s, further assistance in terms of embedded vehicle controlling system will take charge of maneuvering the vehicle and slowly parking on the curb. While parking to the right side of the road, the vehicle control system will continuously check the traffic on the adjacent lane before parking slowly on the curb. Additionally, turn indicators will help maneuver the vehicle by keeping the turn signal on. To model the system, a network is trained using deep learning with 5000 data instances. The trained model is then validated by using real time images from camera to check whether the image of face confirms to the normal pattern or in pain.

In this paper, novel design of a cycloconverter to run a split phase inductor motor to minimize the total harmonic distortion is proposed. The design of a cycloconverter involves two semiconductor switching devices, namely, IGBT and Thyristor for reducing the total harmonic distortion (THD). To realize this, a demand torque and rotor speed of a split phase induction motor is studied. The cycle duration of torque characteristics is divided into suitable number of time intervals (subdivision). These subdivisions of time intervals are in the form of frequencies and used to simplify the cycloconverter design. Change in the frequency at a particular subdivision results in the change in electromagnetic torque of the split phase induction motor. Proposed designs with IGBT and thyristor are compared for their performance with reference to cyclcoconverter designs without any switching devices and with varying firing angles and it is observed that the performance of the thyristor switched cycloconverter reduced the total harmonic distortion more than the IGBT controlled cycloconverter. Varying firing angles were generated using PWM techniques. In other words, split phase induction motor is used as a load for the cycloconverter. However, to drive varying mechanical loads for longer duty cycle, machine needs to minimize the transients which can be made possible with PWM techniques. The output voltage can also be accomplished without any external components for reducing total harmonic distortion (like switching devices). PWM is used to control the switching thereby minimizing the lower order harmonics while it turns on the device and eliminates the higher order harmonics when it turns off the switching device. This gives a reduction in the total harmonic distortion of upto 65.85% for IGBT switching while it reduces to 62.58% with Thyristor controlled switching.

The overall objective of this project was to design a device to collect multiple aerosol samples at various altitudes of the atmosphere. This device would aid in the biological investigation of unexplained microorganisms' travel over vast distances across the world. Extensive work was done in collaboration with the Arkansas State University Biology Department to develop criteria the device needed to meet. The device needed to sample air from sea level up to the outer levels of the Earth's atmosphere. It was determined that the only feasible way to travel to such a distance would be by attaching the device as a payload to the helium weather balloon. Next, various intake and sampling methods were explored. The aim was to collect at least 5 samples varying uniformly in altitude. Beginning at approximately 50,000 feet, a sample would be collected every 10,000 feet until maximum altitude was reached. A centrifugal pump was chosen for intake method. The Allergenco D Impactor, a type of adhesive slide particle capturing container, was chosen for sampling. To prevent contamination, solenoid valves to be used on each side of the impactor to ensure sealing, once the sample was taken. Independent design and testing of all the electrical components for functionality and sensitivity was carried out using a microcontroller to ensure control of valves for sampling at desired altitudes as well as monitoring various parameters. With the use of a vacuum chamber, the sampling system was tested for the required flow rates.

In this paper, a novel watermarking scheme is proposed for super-resolved images. All watermarking methods reduce PSNR of a host signal, while the proposed algorithm increases it. The singular values (SV) derived from edges of the low resolution (LR) image forms the image dependent watermark. Next we obtain linear measurements of the watermark using a compressive sensing (CS). These measurements are embedded in the super-resolved image using a lossless compression framework. Watermark is extracted at the receiver and super-resolved image is authenticated. After positive authentication, watermark (SV's) is used to regenerate edge information which is added to the super-resolved image. This addition increases the PSNR. To best of our knowledge, it is the first algorithm that achieves simultaneous PSNR improvement and provide authentication to the super-resolved image while increasing signal fidelity.

Smart grid is a complex cyber physical system containing numerous and variety of sources, devices, controllers and loads. Communication/Information infrastructure is the backbone of the smart grid system where different grid components are connected with each other through this structure. Therefore, the drawbacks of the information technology related issues are also becoming a part of the smart grid. Further, smart grid is also vulnerable to the grid related disturbances. For such a dynamic system, disturbance and intrusion detection is a paramount issue. This paper presents a Simulink and OPNET based co-simulated platform to carry out a cyber-intrusion in a cyber-network for modern power systems and smart grids. The cyber attack effect is also characterized for the physical power system. The effectiveness of the co simulated platform is demonstrated by the IEEE 30 bus power system model. The distributed denial of service attack was carried out in the cyber network to see its effect on the physical network. Different physical fault situations in the test system are considered and the results indicate the effectiveness of the proposed co-simulated scheme.

A machine learning based model to monitor the smart grid for any suspicious activity or malicious attack is presented in this paper. The model is designed to detect and classify anomalies in the sensory data and is helpful in ensuring the security and stability of the smart grid. The model relies on the real time data collected using wireless sensor networks as an overlay network on the power distribution grid. The overlay network of wireless sensors/devices uses a cluster topology at each tower to collect local information about the tower, and is further augmented by the linear chain topology to connect each tower to the base station (usually at the substation). Preliminary results show that detection mechanism is promising and is able to detect the occurrence of any anomalous event that may cause threat to the smart grid.

This paper addresses the issue of headlight intensity to alleviate glare and blinding during night for drivers. Many factors are considered when analyzing automobile transportation in order to increase safety. One of the most prominent factors for night-time travel is temporary blindness due to elevated headlight intensity. This is particularly prominent on single lane roads. While headlight intensity provides better visual acuity, it inversely affects oncoming traffic. This problem is compounded when both drivers are using a higher headlight intensity setting. Also, higher speed due to decreased traffic levels at night increases the severity of accidents. In order to eliminate accidents due to temporary driver blindness, a fuzzy controller is designed based on the data captured using a wireless sensor network (WSN). Low latency allows quicker headlight intensity adjustment to minimize temporary blindness. Multiple attributes are taken into consideration for controller design. The results show that controller output is nearly instantaneous and generates control signal continuously.

In this paper, hybrid wireless sensor network model is envisaged over the power distribution grid for monitoring the health of the grid. The hybrid model is hierarchical. At the lower level, it uses a cluster topology at each tower to collect local information about the tower while at the higher level it uses linear chain topology to send the grid data to the base station (usually at the substation). Data is collected at each tower, aggregated over the linear chair network, and sent across to a base station for analysis. For analysis, a machine learning based model is employed. The model is designed to detect and classify anomalies in the sensory data and it ensures the security and stability of the smart grid. Initial topology model was investigated using a pilot simulation study followed by experimentation while the analysis is carried using the real time data collected using wireless sensor networks as an overlay network on the power distribution grid. Preliminary results show that detection mechanism is promising and is able to detect the occurrence of any anomalous event that may cause threat to the smart grid.

Superresolution is an algorithmic approach, for constructing high resolution de-noised image from its low resolution and noisier version. A new method to address the problem of copyright violation for super resolution is presented in this paper. The goal is to design an improved watermarking technique, while minimizing distortion in the super resolved image. The approach employs, fuzzy logic to build the perceptual mask, embeds watermark in the low frequency coefficients for robustness with edge preservation and use neural network at the receiver. Novelty lies in providing copyright protection jointly to the low resolution and the super resolved images. The distortion due to watermark insertion is compensated by: 1. use of fuzzy perceptual mask tuned to human visual system; 2. use of trained neural network estimator during watermark extraction; 3. utilize image degradation model during watermark extraction. Effectiveness of the proposed approach is shown by conducting the experiments on natural images and comparing it with the state of the art techniques.

Many factors are considered when analyzing automobile transportation in order to increase safety. One of the most prominent factors for night-time travel is temporary blindness due to elevated headlight intensity. While headlight intensity provides better visual acuity, it inversely affects oncoming traffic. This problem is compounded when both drivers are using a higher headlight intensity setting. Also, higher speed due to decreased traffic levels at night increases the severity of accidents. In order to eliminate accidents due to temporary driver blindness, a wireless sensor network (WSN) based controller is devised to quickly transmit sensor data between cars. Low latency allows quicker headlight intensity adjustment to minimize temporary blindness.

information embedded within them are addressed. Today, a bioinformatics information system typically deals with large data sets reaching a total volume of about one terabyte [25]. Such a system serves many purposes;  User can select the data sources and assign confidence to each selected data source  It organizes existing data to facilitate complex queries  It infers relationships based on the stored data and subsequently predicts missing attribute values and incoming information based on multidimensional data.  Data marts (extension of data warehouse) support different query requests. 2. Data management and integration The Pathway Resource List contains over 150 biological pathway databases and is growing [26]. Usually, first step for the user is to identify a subset of these data sources for integration. To consolidate all the knowledge for a particular organism, extract the pathways from each database need to be extracted and transformed into a standard data representation before integration. Representation of the pathway data in each data source poses another challenge as each pathway modality has its own specific representation issues which must be understood before attempting integration across modalities. For example, metabolic pathways, signal transduction pathways, protein-protein interaction, gene regulation etc. Commonly employed styles of data integration may be implemented in different contexts and under requirements, in order to reuse the data across applications for research collaboration. Some of the data integration and management efforts are presented in [27-32]. Several major approaches have been proposed for data integration, which can be roughly classified into five groups [33-34] namely; data warehousing, federated databasing, serviceoriented integration, semantic integration and wiki-based integration. Across all of these groups, to a significant extent, an increasingly important component of data integration is the community effort in developing a variety of biomedical ontologies to deal in a more specific manner with the technicality and globality of descriptors and identifiers of information that has to be shared and integrated across various resources. Variety of approaches for data integration is discussed below.

Wireless sensor networks (WSN) are proving to be a good fit where real time monitoring of multiple physical parameters is required. In many applications such as structural health monitoring, patient data monitoring, traffic accident monitoring and analysis, sensor networks may involve interface with conventional P2P systems and it is challenging to handle heterogeneous network systems. Heterogeneous deployments will become increasingly prevalent as it allows for systems to seamlessly integrate and interoperate especially when it comes to applications involving monitoring of large infrastructures. Such networks may have wireless sensor network overlaid on a conventional computer network to pick up data from one distant location and carry out the analysis after relaying it over to another distant location.This paper discusses monitoring of bridges using WSN. As a test bed, a heterogeneous network of WSN and conventional P2P together with a combination of sensing devices (including vibration and strain) is to be used on a bridge model. Issues related to condition assessment of the bridge for situations including faults, overloads, etc., as well as analysis of network and system performance will be discussed. When conducted under controlled conditions, this is an important step towards fine tuning the monitoring system for recommendation of permanent mounting of sensors and collecting data that can help in the development of new methods for inspection and evaluation of bridges. The proposed model, design, and issues therein will be discussed, along with its implementation and results.

One of the major challenges of the modern bioinformatics research is to integrate biological pathway data to understand the inner working of the cell. Various pathway data sources are often structured differently and employ algorithms for analysis and integration. Each has a specific motivation for integration that may be suitable only for a particular type of pathway like metabolic pathway or protein-protein interactions. Additionally, with the documentation associated with biological pathway data sources, one needs to understand the database schemas used to store data in each source system, and translate among the schemas in order to exchange information between them. The authenticity of a data source may be subjective as many of them are not independent but derived and data sources often contain similar or overlapping data elements but use conflicting data definitions. There is often a need for user-friendly tools and interfaces to transform bioinformatics data from one database schema to another to discover correlated data among many databases, regardless of the structure of the databases. Most importantly, there are no standards set up for developing biological pathway source and integration. The integration mechanisms may not register important metadata like, copies of input files and time of integration along with the integrated output file. This paper reviews recent developments in biological pathway and sequence data integration and discusses the trends, techniques, issues, and challenges.

Genetic algorithms are robust parallel calculation methods based on natural selection. Various crossover and mutation methods to accomplish Genetic Algorithm (GA), namely, single point, multipoint, uniform, greedy, migration, and on-demand etc.; exist. However, these mechanisms are static in nature. This paper presents a dynamic crossover (DC) mechanism. We investigate its performance by implementing in hardware (FPGA) with convergence rate and higher fitness as the performance metric. The purpose of the DC concept is two fold; to achieve faster convergence and to consume lesser memory by keeping the population size static. The results indicate that for a linear and a nonlinear objective function, DC outperforms all static crossover mechanisms.

Futuristic computers will only be thought of in the context of their ubiquitous connectivity. Net-centric computing isn't communications or networking per se, although it certainly includes both. With the changes in the computing and networking environment we need a different paradigm for distributed computing. The area of net-centric computing encompasses the embedded systems but is much larger in scope. In the near future, many hardware devices will be interconnected in large and highly dynamic distributed systems, using standard communication protocols on standard physical links [1][3][5][6][7][8]. Such types of systems exist only for computers interconnected by TCP/IP networks, or for hardware devices interconnected in small areas by using specific protocols for the physical link, such as Bluetooth, Ethernet or X-10. In this paper we review Net-Centric computing in the perspective of Hardware requirements, Embedded system design, Middleware, Control, IT and provide an insight into the issues and challenges ahead.

Wireless sensor networks consist of a group of nodes, each equipped with sensing, actuating, computation, communication, and storage resources. These sensor nodes are powered by batteries, which are considered as limited resources. Many applications of sensor networks, such as surveillance systems in both civil and military area, habitual monitoring etc., won't allow the replacement of battery supplies. Therefore, to reduce the energy consumption is the key to prolong the lifetime of sensor networks. In this paper, we present two energy efficient data gathering models to achieve longer lifetime in a structured multiclustered topology. The local homogeneous sensor nodes are grouped together to form clusters and a special processing and relaying node is designated to be responsible for communication among local groups. Such models are developed for power transmission line monitoring systems. The goal is to achieve uninterrupted monitoring over a long time using power constrained sensor nodes because the replacement of battery is a major issue in such applications. We use Markov chain process to analyse the proposed two models and comparison shows that the two level communication model consumes less power and is more suitable than single level communication model on the power transmission line monitoring systems