The invention discloses a fatigue state detection method based on a sub-block characteristic matrix algorithm and an SVM (support vector machine), and belongs to the technical field of image processing and mode recognition. The method analyzes and judges whether a driver is in a fatigue state or not through facial features. The method includes the steps: firstly, acquiring a driver video image, and performing illumination compensation and face area detection; secondly, performing eye and mouth area detection in a face area. According to the method, characteristic extraction of an eye image isperformed by an eye sub-block characteristic matrix algorithm, influence of illumination conditions and glasses wearing on detection can be reduced, characteristic extraction of a mouth image is performed by a mouth sub-block characteristic matrix algorithm, interference of tooth appearing and mouth beard in detection can be reduced, images after characteristic extraction are classified by an SVMalgorithm, and reliability is improved under the condition of a small sample training set. According to the method, fatigue characteristics are analyzed according to the eyes and the mouth, the methodtransmits warning information when the driver is in a fatigue state, and traffic accidents can be decreased.