The objective is to use functional near-infrared spectroscopy (fNIRS) to classify and measure mental workload in n-back tasks. Prefrontal cortex (PFC) responses were monitored with fNIRS while 18 subjects completed n-back tasks, performance and NASA-TLX scales were also recorded. It was found that there was significant difference in performance, subjective scores, and fNIRS data under different difficulty tasks; dorsolateral prefrontal cortex (DLPFC) is more sensitive to fNIRS in PFC; using support vector machine (SVM) could realize accurate classification of 4 mental workload levels; SVM decision values could assess and predict mental workload effectively.