This paper presents a portable sensor system to monitor unbalanced head postures of long-distance drivers. The head posture monitoring enables driver fatigue analysis (avoiding potential car accidents) and prevents chronic fatigue and possible neck related diseases. A three-axis MEMS magnetometer and miniature magnet are attached on a user's neck. The user conducts driving scenarios consisting of five common driving actives and one inconvenient posture. Collected magnetometer data are proceeded by neural network algorithms. Part of the data are used to develop a learned model and rest of them are used to calibrate the model. Experiment results are very promising and exhibit superior model accuracy as high as 93.0%.