Incorporating psychoacoustic traits into vibration signal analysis for autonomous squeak noise detection in power window systems

In premium vehicles, even minor noises generated during window operation can significantly influence customers’ perception of quality. Although durability testing is conducted to ensure system reliability, current assessments still depend on manual human auditory evaluation, which introduces potential errors due to subjective judgment and leads to additional energy consumption and operational costs associated with periodic auditory inspections throughout the testing process. To address this limitation, numerous studies have investigated sensor-based approaches for autonomous durability testing. However, existing methods face several practical challenges, including high susceptibility to environmental noise in microphone-based techniques and the requirement for large-scale data and computational resources in deep learning-based approaches, thereby limiting their applicability in real-world durability tests. To overcome these challenges, this study proposes a novel squeak noise detection method for autonomous power window systems utilizing a single vibration sensor. By incorporating signal processing techniques that reflect human auditory perception, the proposed approach enhances the representation of squeak characteristics in vibration signals, enabling robust detection across diverse squeak noise types. The detection capability is evaluated under various squeak occurrence scenarios with varying severity during the durability testing and validated via microphone-based sound inspections. Experimental results demonstrate that the proposed method achieved auditory-label agreement rates of 99% and 98% in two independent vehicle durability tests, thereby confirming its effectiveness and reliability in diagnosing multiple forms of squeak noise during autonomous durability testing in the automotive industry.