Team Introduction

We're a team from makereal labs and department of electronics and electrical engineering in NYCU. Some of us have studied for embedded system and use Jetson Nano, database, and deep learning to make a smart face recognition system. And a member had a lesson about machine learning and deep learning to understand artificial intelligence and build a website which can transfer text into images. For hardware, one of our member make a robot dog by totally himself. From the design of the machine, 3D modeling and programming. Now, we are exert ourself to let the robot dog move.

我們團隊都是交大創客社的隊員。其中有幾位有修習過嵌入式系統的課程，並在期末利用 Jetson Nano 搭配資料庫與深度學習的技術做出智慧人臉系統。也有成員有修過機器學習與深度學習的課程，了解人工智慧的技術，做出從文字生成圖的網站。硬體的部分，我們有隊員自己從零做出一隻機器狗，從機械零件到程式全部都自己動手，現在也著手自己編寫移動系統。

Project Introduction

In the post-epidemic era, the demand for working and studying from home has dramatically increased. On average, every person spends 9.3 hours a day sitting. Sedentary could result in chronic diseases like cardiovascular disease, obesity, and stroke. Bad postures could bring different body pain, including shoulder and neck pain, disc herniation, and arthritis. Therefore, we aim to design a portable smart sitting posture correcting cushion for sedentary people. The built-in sedentary detector reminds users to get enough exercise, thus guiding them to develop a good posture. We focus on developing intelligent chair cushions without the backrest because we want to make the cushion portable and reduce the usage of the sensors. After users sit on the cushion, the pressure sensors on the cushion will send the data to the Raspberry Pi, and Raspberry Pi will scale down the data before transferring it to the ARC EM9D for sitting posture evaluation. Additionally, the vibration motors on the cushion can remind the users of sitting too long. In terms of model training, we will collect data ourselves and train the model with the TensorFlow framework. Afterward, we will transform the model into TensorFlow Lite for edge computing on the ARC board.

身處後疫情時代，人們在家中工作及讀書的需求增高，平均每人每天花9.3小時在椅子上，久坐就像抽菸一樣，會造成心血管疾病、肥胖、中風等慢性病的發生，坐姿不良造成肩頸痠痛、椎間盤突出、關節炎等。因此我們想要為久坐者設計便攜式智慧椅墊，讓使用者能在日常生活中養成良好的坐姿，且設置久坐提醒讓使用者多多活動。
針對龐大的資料處理，我們選擇專注在智慧椅墊的開發而不含靠背等，再資料處理方面使用樹梅派先scaling down後再傳到ARC EM9D 進行坐姿判斷。自行收集資料後用TensorFlow訓練模型，再使用TensorFlow Lite縮小模型空間，減少運算所花的時間。在硬體方面，我們使用壓力感測器及震動馬達，且最小化感測器的運用，製作便利又舒適的椅墊。