Automatický systém pro výstupní kontrolu bezdrátových bluetooth reproduktorů.

Abstract

This bachelor’s thesis focuses on the design and implementation of an automated system for end-of-line testing of wireless Bluetooth speakers. The main objective was to develop a tester capable of detecting defective units based on the analysis of their acoustic response. The proposed system is integrated into a mechatronic workstation with a rotary carousel controlled by a programmable logic controller (PLC), which ensures positioning of tested units and synchronization of the testing sequence. The evaluation of speaker quality is based on a combination of classical audio signal processing methods and machine learning techniques. Mel Frequency Cepstral Coefficients are extracted from the recorded acoustic response and used as input features for a Random Forest classification model, which determines whether a unit is classified as OK or NOK. The thesis also includes the creation of a training dataset, model training, and integration of the trained model into the testing application. The result is a functional demonstration tester enabling automated, objective, and repeatable real-time evaluation of Bluetooth speakers.