A Neural Network Engine for Resource Constrained Embedded Systems

Zuzana Jelcicova, Adrian Mardari, Oskar Andersson, Evangelia Kasapaki, Jens Sparsø

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

41 Downloads (Pure)

Abstract

This paper introduces a dedicated neural network engine developed for resource constrained embedded devices such as hearing aids. It implements a novel dynamic two-step scaling technique for quantizing the activations in order to minimize word size and thereby memory traffic. This technique requires neither computing a scaling factor during training nor expensive hardware for on-the-fly quantization. Memory traffic is further reduced by using a 12-element vectorized multiply-accumulate datapath that supports data-reuse. Using a keyword spotting neural network as benchmark, performance of the neural network engine is compared with an implementation on a typical audio digital signal processor used by Demant in some of its hearing instruments. In general, the neural network engine offers small area as well as low power. It outperforms the digital signal processor and results in significant reduction of, among others, power (5×), memory accesses (5.5×), and memory requirements (3×). Furthermore, the two-step scaling ensures that the engine always executes in a deterministic number of clock cycles for a given neural network.
Original languageEnglish
Title of host publicationProceedings of 54th Asilomar Conference on Signals, Systems, and Computers
PublisherIEEE
Publication date2020
Pages125-131
ISBN (Electronic)978-0-7381-3126-9
DOIs
Publication statusPublished - 2020
Event54th Asilomar Conference on Signals, Systems, and Computers - Pacific Grove, United States
Duration: 1 Nov 20204 Nov 2020

Conference

Conference54th Asilomar Conference on Signals, Systems, and Computers
Country/TerritoryUnited States
CityPacific Grove
Period01/11/202004/11/2020

Fingerprint

Dive into the research topics of 'A Neural Network Engine for Resource Constrained Embedded Systems'. Together they form a unique fingerprint.

Cite this