{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,19]],"date-time":"2026-04-19T22:45:57Z","timestamp":1776638757841,"version":"3.51.2"},"reference-count":40,"publisher":"Wiley","license":[{"start":{"date-parts":[[2019,10,22]],"date-time":"2019-10-22T00:00:00Z","timestamp":1571702400000},"content-version":"unspecified","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["International Journal of Reconfigurable Computing"],"published-print":{"date-parts":[[2019,10,22]]},"abstract":"<jats:p>During the last years, convolutional neural networks have been used for different applications, thanks to their potentiality to carry out tasks by using a reduced number of parameters when compared with other deep learning approaches. However, power consumption and memory footprint constraints, typical of on the edge and portable applications, usually collide with accuracy and latency requirements. For such reasons, commercial hardware accelerators have become popular, thanks to their architecture designed for the inference of general convolutional neural network models. Nevertheless, field-programmable gate arrays represent an interesting perspective since they offer the possibility to implement a hardware architecture tailored to a specific convolutional neural network model, with promising results in terms of latency and power consumption. In this article, we propose a full on-chip field-programmable gate array hardware accelerator for a separable convolutional neural network, which was designed for a keyword spotting application. We started from the model implemented in a previous work for the Intel Movidius Neural Compute Stick. For our goals, we appropriately quantized such a model through a bit-true simulation, and we realized a dedicated architecture exclusively using on-chip memories. A benchmark comparing the results on different field-programmable gate array families by Xilinx and Intel with the implementation on the Neural Compute Stick was realized. The analysis shows that better inference time and energy per inference results can be obtained with comparable accuracy at expenses of a higher design effort and development time through the FPGA solution.<\/jats:p>","DOI":"10.1155\/2019\/7218758","type":"journal-article","created":{"date-parts":[[2019,10,22]],"date-time":"2019-10-22T23:30:41Z","timestamp":1571787041000},"page":"1-13","source":"Crossref","is-referenced-by-count":45,"title":["An FPGA-Based Hardware Accelerator for CNNs Using On-Chip Memories Only: Design and Benchmarking with Intel Movidius Neural Compute Stick"],"prefix":"10.1155","volume":"2019","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-0123-7977","authenticated-orcid":true,"given":"Gianmarco","family":"Dinelli","sequence":"first","affiliation":[{"name":"Department of Information Engineering, University of Pisa, Pisa 56122, Italy"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9311-6392","authenticated-orcid":true,"given":"Gabriele","family":"Meoni","sequence":"additional","affiliation":[{"name":"Department of Information Engineering, University of Pisa, Pisa 56122, Italy"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4889-0968","authenticated-orcid":true,"given":"Emilio","family":"Rapuano","sequence":"additional","affiliation":[{"name":"Department of Information Engineering, University of Pisa, Pisa 56122, Italy"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1723-7421","authenticated-orcid":true,"given":"Gionata","family":"Benelli","sequence":"additional","affiliation":[{"name":"IngeniArs, Pisa 56121, Italy"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5426-4974","authenticated-orcid":true,"given":"Luca","family":"Fanucci","sequence":"additional","affiliation":[{"name":"Department of Information Engineering, University of Pisa, Pisa 56122, Italy"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"311","reference":[{"key":"1","doi-asserted-by":"publisher","DOI":"10.1007\/s10489-013-0461-5"},{"issue":"8","key":"2","first-page":"1289","article-title":"Advances and perspective on applications of deep learning in visual object detection","volume":"43","author":"H. Zhang","year":"2017","journal-title":"Acta Automatica Sinica"},{"key":"3","doi-asserted-by":"publisher","DOI":"10.1016\/j.neucom.2017.02.016"},{"key":"4","doi-asserted-by":"publisher","DOI":"10.1007\/s00521-012-0892-7"},{"key":"5","first-page":"13","article-title":"Exploring critical aspect of CNN-based keyword spotting. A phocnet study","author":"G. Retsinas"},{"key":"6","first-page":"285","article-title":"Keyword spotting using support vector machines","author":"Y. B. Ayed"},{"key":"7","article-title":"Hello edge: keyword spotting on microcontrollers","author":"Y. Zhang","year":"2017"},{"key":"8","doi-asserted-by":"publisher","DOI":"10.1016\/j.neucom.2018.09.038"},{"key":"9"},{"key":"10"},{"key":"11","doi-asserted-by":"publisher","DOI":"10.1007\/s00521-018-3761-1"},{"key":"12","doi-asserted-by":"crossref","article-title":"A low power keyword spotting algorithm for memory constrained embedded system","author":"G. Benelli","DOI":"10.1109\/VLSI-SoC.2018.8644728"},{"key":"13"},{"key":"14","unstructured":"MohamedA.Deep neural network acoustic models for ASR2014Toronto, CanadaToronto UniversityDoctoral thesis"},{"key":"15","first-page":"448","article-title":"Batch normalization: accelerating deep network training by reducing internal covariate shift","author":"S. Ioffe"},{"key":"16","first-page":"161","article-title":"Optimizing FPGA-based accelerator design for deep convolutional neural network","author":"C. Zhang"},{"key":"17","first-page":"26","article-title":"Going deeper with embedded FPGA platform for convolutional neural network","author":"J. Qiu"},{"key":"18","first-page":"77","article-title":"Accelerating binarized neural networks: comparison of FPGA, CPU, GPU and ASIC","author":"E. Nurvitadhi"},{"key":"19","doi-asserted-by":"publisher","DOI":"10.1109\/tcsii.2017.2690919"},{"key":"20","first-page":"1","article-title":"Accelerating low bit-width convolutional neural networks with embedded FPGA","author":"L. Jiao"},{"key":"21","first-page":"1011","article-title":"FPGA based implementation of deep neural networks using on-chip memory only","author":"J. Park"},{"key":"22","doi-asserted-by":"publisher","DOI":"10.1109\/jproc.2017.2761740"},{"key":"23","first-page":"624","article-title":"Energy-efficient and high-throughput FPGA-based accelerator for convolutional neural networks","author":"G. Feng"},{"key":"24","first-page":"1","article-title":"High-performance video content recognition with long-term recurrent convolutional network for FPGA","author":"X. Zhang"},{"key":"25","first-page":"268","article-title":"FPGA-based CNN inference accelerator synthesized from multi-threaded C software","author":"J. H. Kim"},{"key":"26"},{"key":"27"},{"key":"28"},{"key":"29"},{"key":"30"},{"key":"31"},{"key":"32"},{"key":"33"},{"key":"34"},{"key":"35"},{"key":"36"},{"key":"37","article-title":"Handwritten digit recognition based on depth neural network","author":"Y. Hout"},{"key":"38","doi-asserted-by":"crossref","article-title":"Transfer learning for Latin and Chinese characters with deep neural networks","author":"D. C. Ciresan","DOI":"10.1109\/IJCNN.2012.6252544"},{"key":"39","article-title":"Very deep multilingual convolutional neural networks for LVCSR","author":"T. Secu"},{"key":"40","article-title":"MobileNets: efficient convolutional neural networks for mobile vision applications","author":"A. G. Howard","year":"2017"}],"container-title":["International Journal of Reconfigurable Computing"],"original-title":[],"language":"en","link":[{"URL":"http:\/\/downloads.hindawi.com\/journals\/ijrc\/2019\/7218758.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"http:\/\/downloads.hindawi.com\/journals\/ijrc\/2019\/7218758.xml","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"http:\/\/downloads.hindawi.com\/journals\/ijrc\/2019\/7218758.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2021,1,12]],"date-time":"2021-01-12T20:21:05Z","timestamp":1610482865000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.hindawi.com\/journals\/ijrc\/2019\/7218758\/"}},"subtitle":[],"editor":[{"given":"Martin","family":"Margala","sequence":"additional","affiliation":[],"role":[{"role":"editor","vocabulary":"crossref"}]}],"short-title":[],"issued":{"date-parts":[[2019,10,22]]},"references-count":40,"alternative-id":["7218758","7218758"],"URL":"https:\/\/doi.org\/10.1155\/2019\/7218758","relation":{},"ISSN":["1687-7209","1687-7195"],"issn-type":[{"value":"1687-7209","type":"electronic"},{"value":"1687-7195","type":"print"}],"subject":[],"published":{"date-parts":[[2019,10,22]]}}}