在本論文中，我們發展了一個以慣性感測器為基礎的汽車駕駛行為分析系統。此系統可以讓我們偵測出車輛的加速、減速、左轉、與右轉等事件狀況是否為正常行駛或是激烈行駛；另外，我們也可以偵測車輛行駛中的路面是否過於顛簸。 我們使用了開放式軟硬體平台Arduino作為運算的核心，並搭配了三軸加速度計與三軸陀螺儀等慣性感測元件作為分析信號的來源。我們在感測信號的前置處理上，除了事先校正量測誤差之外，也採用了數位低通濾波器以濾除一些車輛引擎或路面所帶來的震動干擾。 為了能夠更可靠的偵測出駕駛行為的多種事件，使用了以模糊邏輯理論做為分析基礎的判斷方法。模糊邏輯包含了梯形歸屬函數模糊化、最大最小合成法、及重心解模糊等主要步驟。經過以上步驟後，我們最後依據各邏輯判斷的結果得到一個正確的駕駛事件分類。 最後實際於車輛上的實驗中，以2個人做為乘客，分別紀錄乘坐在車內之車輛行駛狀況；再與模糊邏輯的駕駛行為分析系統所產生的事件比對，驗證了我們的系統確實可以成功的偵測到各種不同的駕駛行為事件，其判定的結果也與乘客的認知接近。 ;In this thesis, we have developed an inertial sensor-based automobile driver behavior analysis system. This system can help us to detect if a car is in a normal or extreme driving condition during vehicle acceleration, deceleration, and left or right turning. We used an Arduino open hardware and software platform core, and a three-axis accelerometer and three-axis gyroscope inertial sensing element analysis as a source of the signal. In the pre-processing of the sensed signals we used a digital low pass filter to filter out some of the vehicle engine or road surface interference caused by vibration. This was done in addition to previous measurement error correction. To be able to more reliably detect a variety of driving behavior events, we used the fuzzy logic theory as the basis of our analytic judgment. Fuzzy logic includes fuzzy membership function, the main step synthesis, and the maximum and minimum gravity defuzzification. After the above steps, we finally got a proper driving event classification based on the results of each logic judgment. Finally, we conduct experiments on a vehicle. Two passengers in a running vehicle record the vehicle status sequences. The status sequences were compared with those generated by the proposed behavior analysis system based on the fuzzy logic theory. The experiments results validate that indeed the system can successfully detect various driving behavior events; the results generated by the proposed system are consistent with the determination of cognitive passengers.