SIMO: An Automatic Speech Recognition System for Paperless Manufactures

Abstract

Despite environmental general conscience, heavy use of paper is still one fact in nowadays factories. The shorter the manufacturing production, the greater the tendency to employ paper to support quality tracking of pieces; using it to register measurements or nonconformities. This tendency increases drastically in some manufactures like aerospace, where typical production ratios vary between 9 and 18 subassemblies per month. The current work presents an automatic speech recognition system, meant to replace paper by a digitalized version of the manual writing task. The work presents (i) industrial use cases with benefits and requirements; (ii) the system architecture, including several tested free Automatic Speech Recognition modules, their analysis; and (iii) some open-source supporting modules that improves its functionality. The work concludes presenting several tests, showing the system performance against different kind of industrial noises, low to high quality microphones and users with different dialects.