Voiceprint Testing Machine

The voiceprint testing system was independently developed by our company to address the issue of poor insertion and removal of battery latch hooks in automotive lithium batteries. It is an intelligent testing system developed based on differences in detected sounds, capable of detecting specific voiceprint signals under various sound environments. By judging the difference between detected sounds and normal sounds, it identifies product quality, and can be used in fault detection, voiceprint classification, and various industrial processes that can be distinguished and classified by sound. Using this system enables efficient, automated detection of whether voiceprints are normal, thereby carrying out corresponding operations. To achieve better detection accuracy and performance, the current system is equipped with AP series high-precision acoustic analyzers, supporting multiple sampling rates such as 48K, 96K, and 192K. AP series products have gained widespread recognition in the industry for their low system noise floor, high precision, and good stability. Microphones and microphone power supplies all adopt the GRAS brand, with excellent performance and strong product competitiveness.
The software part is independently developed by our company's software team, with strong architecture robustness. The multi-threaded approach ensures that software interface response, software timing, signal acquisition, PLC communication, and other tasks can be performed independently and simultaneously, and all tasks can start and stop simultaneously, safely and efficiently. The software has built-in Modbus TCP/IP communication, supports both single-step operation and automatic operation dual modes, with adjustable sampling rate and sampling time, making user use more flexible, and supports custom settings and judgment of up to 50 different sound types. The algorithm part combines cutting-edge algorithms in related fields and feature analysis of instantaneous pulse signals. After extensive experiments and simulation research, it ultimately provides users with five high-resolution features, which on one hand can be set with upper and lower limits according to on-site conditions, and on the other hand can also enable only some of the features for different combinations, thereby improving recognition accuracy and robustness.​