-
High Speed SRT Divider for Intelligent Embedded System
Authors:
Bhavana Mehta,
Jonti Talukdar,
Sachin Gajjar
Abstract:
Increasing development in embedded systems, VLSI and processor design have given rise to increased demands from the system in terms of power, speed, area, throughput etc. Most of the sophisticated embedded system applications consist of processors, which now need an arithmetic unit with the ability to execute complex division operations with maximum efficiency. Hence the speed of the arithmetic un…
▽ More
Increasing development in embedded systems, VLSI and processor design have given rise to increased demands from the system in terms of power, speed, area, throughput etc. Most of the sophisticated embedded system applications consist of processors, which now need an arithmetic unit with the ability to execute complex division operations with maximum efficiency. Hence the speed of the arithmetic unit is critically dependent on division operation. Most of the dividers use the SRT division algorithm for division. In IoT and other embedded applications, typically radix 2 and radix 4 division algorithms are used. The proposed algorithm lies on parallel execution of various steps so as to reduce time critical path, use fuzzy logic to solve the overlap problem in quotient selection, hence reducing maximum delay and increasing the accuracy. Every logical circuit has a maximum delay on which the timing of the circuit is dependent and the path, causing the maximum delay is known as the critical path. Our approach uses the previous SRT algorithm methods to make a highly parallel pipelined design and use Mamdani model to determine a solution to the overlap** problem to reduce the overall execution time of radix 4 SRT division on 64 bits double precision floating point numbers to 281ns. The design is made using Bluespec System Verilog, synthesized and simulated using Vivado v.2016.1 and implemented on Xilinx Virtex UltraScale FPGA board.
△ Less
Submitted 17 February, 2018;
originally announced February 2018.
-
Implementation of SNR estimation based Energy Detection on USRP and GNU Radio for Cognitive Radio Networks
Authors:
Jonti Talukdar,
Bhavana Mehta,
Kinjal Aggrawal,
Mansi Kamani
Abstract:
Development of smart spectrum sensing techniques is the most important task in the design of a cognitive radio system which uses the available spectrum efficiently. The adaptive SNR estimation based energy detection technique has the dual benefit of improving the efficiency of spectrum usage by capitalizing on the underutilization of the spectrum in an adaptive and iterative fashion, as well as re…
▽ More
Development of smart spectrum sensing techniques is the most important task in the design of a cognitive radio system which uses the available spectrum efficiently. The adaptive SNR estimation based energy detection technique has the dual benefit of improving the efficiency of spectrum usage by capitalizing on the underutilization of the spectrum in an adaptive and iterative fashion, as well as reducing the hardware resources leading to easy implementation on a versatile and diverse group of cognitive radio infrastructures. The use of adaptive threshold for energy detection based on SNR estimation improves the spectrum sensing performance and efficiency of the cognitive radio by many folds, especially in low SNR as well as high noise variance situations. The proposed method is implemented on the USRP B200 and results show significant improvement in the detection rate of primary users as compared to conventional energy detection techniques.
△ Less
Submitted 8 November, 2017; v1 submitted 22 August, 2017;
originally announced August 2017.
-
Human Action Recognition System using Good Features and Multilayer Perceptron Network
Authors:
Jonti Talukdar,
Bhavana Mehta
Abstract:
Human action recognition involves the characterization of human actions through the automated analysis of video data and is integral in the development of smart computer vision systems. However, several challenges like dynamic backgrounds, camera stabilization, complex actions, occlusions etc. make action recognition in a real time and robust fashion difficult. Several complex approaches exist but…
▽ More
Human action recognition involves the characterization of human actions through the automated analysis of video data and is integral in the development of smart computer vision systems. However, several challenges like dynamic backgrounds, camera stabilization, complex actions, occlusions etc. make action recognition in a real time and robust fashion difficult. Several complex approaches exist but are computationally intensive. This paper presents a novel approach of using a combination of good features along with iterative optical flow algorithm to compute feature vectors which are classified using a multilayer perceptron (MLP) network. The use of multiple features for motion descriptors enhances the quality of tracking. Resilient backpropagation algorithm is used for training the feedforward neural network reducing the learning time. The overall system accuracy is improved by optimizing the various parameters of the multilayer perceptron network.
△ Less
Submitted 22 August, 2017;
originally announced August 2017.
