探討Media-Pipe和Leap Motion藉由用於發展遲緩的VR系統在分層共現網路的對指手勢預測精準度比較;Using Hierarchical Co-occurrence Network to Compare the Prediction Accuracy of Fine Motor Gesture between Media-Pipe and Leap Motion via a VR System for Developmental Delays

NCUIR > College of Electrical Engineering & Computer Science > Software Engineer > Electronic Thesis & Dissertation > Item 987654321/89638

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/89638

Title:	探討Media-Pipe和Leap Motion藉由用於發展遲緩的VR系統在分層共現網路的對指手勢預測精準度比較;Using Hierarchical Co-occurrence Network to Compare the Prediction Accuracy of Fine Motor Gesture between Media-Pipe and Leap Motion via a VR System for Developmental Delays
Authors:	游盛棋;Yu, Sheng-Chi
Contributors:	軟體工程研究所
Keywords:	早期療育;精細動作;分層共現網路;轉移學習;監督式學習;media pipe;AIOT;leap motion;Early intervention;fine motor assessment;hierarchical co-occurrence network;transfer learning;supervised learning;media pipe;AIOT;leap motion
Date:	2022-08-30
Issue Date:	2022-10-04 11:50:20 (UTC+8)
Publisher:	國立中央大學
Abstract:	發展遲緩兒童的診斷以及治療需要有專業人員的協助才能進行診斷並規劃治療流程，並且需要隨時篩查兒童的狀況以擬訂不同的治療方針，但在偏鄉缺乏人力及資源的情況下，孩童的障礙特徵容易被忽視，因而錯失早期介入的黃金時期。我們開發了一個系統，可以藉由遠端的方式，由診療師給定任務，孩童完成任務後，遊戲會回傳遊玩結果給雲端資料庫，該雲端資料庫將數據處理為可視化的型式傳給診療師，以利後續追蹤及任務難度修改。在對指的手勢識別上，我們使用hierarchical co-occurrence network (HCN)分類手勢，該架構結合空間資訊和時間資訊的輸入，以達到全域共現的效果。在過去，我們有收集LMC在抓木塊遊戲的數據，因此嘗試利用mapping model，將LMC的數據映射為media pipe的數據，協助我們的遊戲在AIOT的應用。;The diagnosis and treatment of children with developmental delays requires professional assistance to diagnose and plan the treatment process. Child’s condition needs to be screened at all times to develop a different treatment approach. However, in rural areas, where there is a lack of resources, the characteristics of the child′s impairment can easily be overlooked, thus missing a golden opportunity for early intervention. We have developed a system where tasks are given by the therapist via a remote location and when the child completes the task, the game sends back the results to a cloud-based database which processes the data into a visual format for follow-up and task modification. For finger gesture recognition and developmental recognition task, we use a hierarchical co-occurrence network to classify gestures, which combines the input of spatial and temporal information to achieve global co-presence. In the past, we have collected data from LMC in grab wood block game, so we try to use the mapping model to map the data from LMC to media pipe to help our game in artificial intelligence of things (AIOT) application.
Appears in Collections:	[Software Engineer] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	148	View/Open

社群 sharing

Loading...