Title: Performance Analysis of Filter Bank Multicarrier based Communication System
Authors: A Navya and D. Kiran kumar
Abstract: Recently two different L-band digital aeronautical with L-DACS1 selected as the best candidate. In this paper we describe a filter bank multicarrier (FBMC) based communication system, and show its advantages over the LDACS systems. We provide simulation results for all three communication systems to fairly compare their power spectral density (PSD), peak-to-average power ratio (PAPR), and BER performance. We show that in a measurement-based AG communication channel model, FBMC has better performance (and spectral containment) than the L-DACS schemes, and this is particularly true in the presence of actual interfering signals from distance measuring equipment (DME). Simulation results show that FBMC can substantially reduce the out-of-band (OOB) power, and can suppress DME interference by at least 19.5 dB, due to its well-localized subcarrier prototype filters. FBMC can also increase throughput and spectral efficiency by reducing the number of guard-band subcarriers and removal of the cyclic prefix, postfix and windowing techniques used in L-DACS1.These results show that an FBMC based communication system can be an appealing candidate for future AG communication systems.
Title: Multiple Human Action Recognition using Big Data Set
Authors: Vandna and Krishan Kumar
Abstract: Motivation that urges this investigation is to pick whether the video dataset presently has enough realities to plot a profound convolutional tactile framework (CNNS) with 3-dimensional (three-D) spatio-appearance. Till the stop, the three-D CNN show charge inside the standing of activity has been extended in any way. Around then, hitherto, standard examinations checked through 3-d setups turned out to be generally shallow. We broke the different CNN 3-D constructions for most extreme shallow profundities on the current video dataset. Given the impacts of the investigate, venture through conclusive touch can be gotten: (I) ResNet-18 plans over fitting for UCF-101, major human development information base (HMDB51), along the edge of dynamic side interest (alright). ii.) Kinetics (k) dataset has the appropriate records to get outfitted from CNN bubbling 3-D, and licenses RESNET getting prepared however much 152 layers, similar as a 2-D resnet (size) in imageNET. ResNet-101 arrives at seventy eight.4 percent exactness in an average energy check set. (iii) Kinetics model 3D Pretrainly Beat confounded shape second and Pretrained ResNet-one zero one arrived at ninety four.5 rate and 70.2 rate at UCF-one zero one and HMDB51, detached. Utilizing 2d CNNS Systematize on ImageNET has proficient enormous advancement in different obligations inside the photo.
Title: Cascaded Integrator Comb Filter Design for Digital Down Converter on FPGA
Authors: P.A.Harsha Vardhini and P.Sandeep
Abstract: A digitized real signal centered at intermediate frequency is converted by digital down converter to a signal centered at zero frequency. The multiplications of quadrature mixing are done at much lower rate of sampling instead of high sample rates. When there is a rate conversion of signal to a new clock frequency then there is a need of high clock rate filter architectures. At such high frequency, FIR filter variants fail to work. Cascaded Integrator Comb filters are the best choice at these high sampling rates as they support for large decimation reducing the complexity & consumption of power. CIC architecture is realized using VHDL and is simulated on Xilinx FPGA.