Edge Intelligence Empowered English Speech Recognition in Portable Devices

ABSTRACT Online education has greatly promoted English learning. However, many learners still meet difficulty in spoken English. With the fast development of computer science and artificial intelligence, especially natural language processing, online education is becoming more and more intelligent which can greatly assist learners to grasp English speaking skill thorough automatically English speech recognition. However, it is still a challenge task to implement English speech recognition in portable devices which has limited computing resources. In order to tackle this issue, this paper proposes to combine edge intelligence with cloud computing for mobile English speech recognition. First, the preprocessing and feature extraction part is deploy in the edge nodes, such as pad or smart phone. Second, a pretrained speech recognition model is deployed in the cloud server to avoid the limits of edge nodes. The features are sent from edge nodes to cloud server, while the pretrained speech model in the cloud server returns the results to edge nodes. The experiments and simulations demonstrate the proposed solution can adapt the edge nodes with limited resources.