User identification and personalized interaction of smart desk lamps by fusing convolutional LSTM networks and conditional generative adversarial networks

Abstract

Listen

Translate article

ABSTRACT In the rapid development of smart homes, various kinds of smart furniture have also been developed rapidly. This research aims to fuse convolutional LSTM network and conditional generative adversarial network to achieve user identification and personalized interaction of smart desk lamps. First, the user’s facial features are extracted using a convolutional neural network for identity recognition. Then, the extracted facial features are used as conditional inputs, which are combined with the generator network to generate personalized interaction content through the conditional generative adversarial network. Results showed that the anomaly scores during the operation of the model are all lower than 0.3%, and the loss rate of the model is basically controlled within 0.4%. In addition, the average accuracy of the proposed model reaches 94.86% and the average recall rate reaches 92.74%, while the image output time is 2.44 s and 2.15 s on the two datasets, which is much lower than that of other control models. After recognizing the user’s identity, the smart desk lamp can provide personalized services and interactions to meet the user’s needs and preferences. This research provides an effective solution for user identification and personalized interaction in the smart home domain.