利用多媒體處理技術的智慧型自走車在移動目標物的跟隨控制設計

本論文的主要重點是設計一部可以應用於日常生活中的智慧型自走車，此自走車能夠利用結合影像與語音的多媒體辨識技術，來達成目標物的跟隨控制目的，除了能夠提升使用上的多功能性應用，更能達到降低人力成本以及增進人們生活更便利的目標。此系統主要可分為兩個部分:多媒體處理與自走車設計。在多媒體處理的部分，我們使用了影像處理及語音辨識技術，利用Kinect 和語音辨識晶片(ASR-M08A) 來獲得所有包括影像及語音的相關資訊。我們利用Kinect的鏡頭來獲取彩色及深度的資訊，然後使用影像處理來得到我們需要的骨架資訊。另外，我們利用Kinect 的麥克風陣列和語音辨識晶片獲得語音資訊，經過語音辨識和聲源偵測所獲取的資訊，我們就能透過聲音讓自走車辨識語音指令和使用者的方位。在自走車設計的部分，我們的系統使用PID控制和模糊控制，來操縱車子的移動方向及速度。我們依據目標物所在的位置和Kinect感測到目標物與車的距離，來建立自走車的左輪和右輪的模糊控制規則庫，車子就能依照我們的規則庫做出相對應的輸出，進而達成跟隨控制的目標。在系統啟動後，自走車根據使用者的語音命令來判斷目標位置再去鎖定目標，然後再使用骨架辨識繼續跟隨目標物，而使用語音辨識的功能就可以更直覺的控制自走車，更能有效的避免自走車找不到目標使用者的問題。而在自走車在跟隨的過程中若是遭遇障礙物，系統中使用的紅外線感測器就能用來感測障礙物的距離並避開，因此自走車能在複雜環境中成功地達到目標物跟隨控制的目的。

關鍵字

智慧型自走車；語音辨識； Kinect ；模糊控制；骨架辨識追蹤

並列摘要

The main purpose of this thesis is to design an intelligent vehicle that can be applied to our daily lives. This intelligent vehicle can use multimedia processing with both image and speech to achieve the purpose of target following control. In addition to enhancing the versatility of vehicle application, it can achieve the goal of reducing labor costs and helping people to live more conveniently. This system can be roughly divided into two parts, including multimedia processing and intelligent vehicle design. In the section of multimedia processing, we employ Kinect and speech recognition chip (ASR-M08A) to obtain all helpful information, including image and voice. We use Kinect camera to obtain color and depth information, which would be applied for some image processing to obtain the skeleton information. In addition, we use Kinect's microphone array and speech recognition chip to obtain voice information. With speech recognition and sound source detection, the vehicle is able to identify speech commands and user orientation. In the part of intelligent vehicle design, our system uses PID control and fuzzy control to manipulate the direction and speed of movement of the vehicle. According to the target location and the Kinect sensed distance between people and vehicle, the fuzzy control rule bases can be built for the left and right wheels. Therefore, the vehicle would able to obey our fuzzy rules to yield the corresponding output for the purpose of target following. After the system starts, it can employ the user's speech commands to determine the target location and then lock the target. This system uses the skeleton identification to continue to follow the target. The use of speech recognition can be more intuitive to control the behavior of vehicle and more effective to restore the target with lost target problem. If there is an obstacle in front of the vehicle, the infrared sensors used in the system can sense the location of the obstacle and avoid hitting it, so the intelligent vehicle can successfully achieve the target following purpose in a complicated environment.