Loading [a11y]/accessibility-menu.js
A Comprehensive Review of ML-based Time-Series and Signal Processing Techniques and their Hardware Implementations | IEEE Conference Publication | IEEE Xplore

A Comprehensive Review of ML-based Time-Series and Signal Processing Techniques and their Hardware Implementations


Abstract:

Advancements in technology and smart devices demand incubation of advanced learning methodologies to learn or capture the underlying patterns in the data and further util...Show More

Abstract:

Advancements in technology and smart devices demand incubation of advanced learning methodologies to learn or capture the underlying patterns in the data and further utilize and analyze the extracted patterns for different purposes such as anomaly detection, face recognition, and data prediction. There are many techniques for learning and analyzing data patterns, each of which has different complexity and performance. Two important kinds of data analyzed in the current day technology are the time-series signals and the images (computer vision). In this work, we outline diverse machine learning techniques and their applicability for pattern analysis in the aforementioned kind of applications. This survey is motivated by the increasing need to analyze time-series signals and image processing in resource-constrained, embedded devices for a large variety of applications, including autonomous driving. First, we provide a comprehensive review of techniques for analyzing the signals based on numerous stochastic, traditional machine learning, deep neural networks, and other hybrid approaches. For all the prominent families of techniques, we discuss their use in prediction, anomaly detection, and classification. Moreover, we consider their algorithmic complexity and review their hardware implementations. Further, we compare and analyze the involved trade-offs among different techniques (including hardware requirements), and its suitability for analyzing various data sets such as image classification, recognition, and time-series analysis.
Date of Conference: 19-22 October 2020
Date Added to IEEE Xplore: 28 December 2020
ISBN Information:
Conference Location: Pullman, WA, USA

References

References is not available for this document.