Optimizing the Human-Computer Interaction Interface of Warehouse Management Systems Using Automatic Speech Recognition Technology

The exponential growth of the e-commerce and logistics industries in recent years has underscored the necessity for a warehouse management system (WMS) that is more efficient and intelligent. The prevailing WMSs are dependent on manual input and the use of handheld devices, which can result in inefficiencies and the potential for human error. In this work, we propose a human-computer interaction model for speech recognition optimized for the warehouse environment. This model integrates noise suppression technology based on an adaptive filter, which can dynamically detect and filter background noise, such as forklift operation sounds, human voices, and mechanical operation sounds. Furthermore, to address the variability in pronunciation, speech rate, and accent among different users, the system incorporates a speech enhancement model based on the variational autoencoder (V AE) technique. This approach enables the system to adaptively adjust the input speech features, thereby enhancing the robustness of recognition. In regard to natural language processing, a natural language understanding module based on bidirectional encoder representations from transformers (BERT) has been developed. The module is capable of semantic parsing of instructions in the user's voice and converting them into executable operation commands. Semantic slot filling technology enables the system to automatically identify the key entities in a task and perform linkage operations with the backend WMS database. The experimental analysis demonstrates that the proposed system is effective in an actual warehouse scenario. Compared with the traditional method, the task completion speed and accuracy are significantly improved.