Showing papers by "MKSSS's Cummins College of Engineering for Women published in 2022"
TL;DR: In this paper, the implications of applying silver nanoparticles in agriculture and their possible consequences are comprehensively discussed, and the potential of biogenic nanoparticles-viable antimicrobial agents for enhanced applications in agriculture as biopesticides are also evaluated.
38 citations
01 Jan 2022
TL;DR: In this paper, a You Only Look Once (YOLO) on a robotic environment working on the number of clients and servers ends is presented, where output efficiency leverage is seen with AI acceleration with toolkits like Intel's OpenVINO.
Abstract: With the development of cloud robotics, a much broader scope of multidisciplinary applications to create smart systems is now available. The “Artificially Intelligent” system's brains are in the cloud. The cloud can hold data centers, deep learning, communication support, etc. With the help of edge computing, VM-based cloudlets, deploying deep learning implementation systems are a more practical option rather than one single system doing all the tasks. The mobile applications and IoT devices often produce streaming data which requires real-time analysis and control. When the application involves end devices as hardware like Raspberry Pi and laptops working at edge, an acceleration in the result generation is also necessary. This paper presents its observations with the implementation of one such machine learning application of object detection and recognition, i.e., You-Only-Look-Once (YOLO) on a robotic environment working on the number of clients and servers ends. Differentiating cloud and edge, we have demonstrated the analysis and results where output efficiency leverage is seen with AI acceleration with toolkits like Intel's OpenVINO.
01 Jan 2022
TL;DR: In this article, a particle swarm optimization-based noise estimation for single-channel speech enhancement was proposed. But, the proposed algorithm converges fast while giving an optimal solution, and the perceptual estimate of speech quality (PESQ) of the enhanced speech with the results of various other algorithms.
Abstract: Background noise affects speech quality and intelligibility degrading the performance of speech-operated systems. Speech enhancement can improve the quality of noisy speech. Modulation domain spectral subtraction for separate real and imaginary spectra improves speech intelligibility without having musical noise in enhanced speech. In this paper, we suggest using particle swarm optimization-based noise estimation for single-channel speech enhancement. Particle swarm optimization algorithm finds the most optimum value of noise present in input speech. We investigate the suitability of this algorithm for noise estimation in the modulation domain by comparing the perceptual estimate of speech quality (PESQ) of the enhanced speech with the results of various other algorithms. The proposed algorithm converges fast while giving an optimal solution. We get improvement in PESQ and segmental SNR values. The musical noise is reduced in enhanced speech.