Hardware-Accelerated, Short-Term Processing Voice and Nonvoice Sound Recognitions for Electric Equipment Control

We proposed and implemented a sound recognitionsystem for electric equipmentcontrol. In recent years, industry 4.0 has propelled a rapid growth in intelligent human–machine interactions. User acoustic voice commands for machine controlhave been examined the most by researchers. The targeted machine can be controlled through voice without the use of any hand-held device. However, compared with human voicerecognition, limited research has been conducted on nonhuman voice (e.g., mewingsounds) or nonvoice sound recognition(e.g., clapping). Processing of such short-term, biometric nonvoice sounds for electric equipmentcontrol requires a rapid response with correct recognition. In practice, this could lead to a trade-off between recognition accuracy and processing performance for conventional software-based implementations. Therefore, we realized a field-programmable gate array-based embedded system, such a hardware-acceleratedplatform, can enhance information processing performance using a dynamic time warpingaccelerator. Furthermore, information processing was refined for two specific applications (i.e., mewingsounds and clapping) to enhance system performance including recognition accuracy and execution speed. Performance analyses and demonstrations on real products were conducted to validate the proposed system.