WEARABLE DEVICES TO ASSIST THE VISUALLY IMPAIRED

Abstract

The invention proposes a personal assistant device based on a Vision-Language Model (VLM) to assist visually impaired people in daily activities. The device includes: (i) a wearable assistant device (100) consisting of a camera (113), microphone (114), push button (115) and speaker (116); (ii) a processing unit (120) executing an interactive control block (140) consisting of a state machine (141), a multimodal data buffer (142) and an AI model connection interface (143); and (iii) an artificial intelligence service layer (150) consisting of a speech recognition service (151), a vision-language model service (152) and a speech synthesis service (153).The method includes the following steps: capturing audio data when the user presses and holds the button (115) (S100); capturing a still image when the user releases the button (115) (S200); linking audio and still image data in the multimodal data buffer (142) (S300); sending a processing request to the artificial intelligence service layer (150) (S400); analyzing the combination of voice and image to create a semantic response (S500); and playing the audio response through the speaker (116) and then putting the device into standby mode (S600).