Error detection by sound recognition in automated pipetting

Our project details the development of a sound recognition algorithm used for detecting errors in automated pipetting to insure continues quality control. The system was developed in cooperation with TechVolver Aps, where we tested the algorithm throughout the development process. By making sound signatures of a pipetting process, meaning a range of amplitude and frequency in which the pipetting normally occurs, we can determine whether a new pipetting is with or without error. We analyzed the construction of the pipette holder to insure that only the sound from the pipette would be recorded by the microphone attached to it. TechVolver estimates that the method has the potential to reduce errors by up to 80%, and reduce pipette inspection by up to 70%.

Category: ENGINEERING Country: DENMARK Year: 2021


Filip Kikkenborg Kikkenborg

Thorbjørn Valdemar Ræder Clausen

Asger Ren Nordbjerg