In order to investigate the feasibility of mental workload (MWL) assessment influenced by different emotional state by using the functional near-infrared spectroscopy (fNIRS) and to support the operator functional state (OFS) assessment based on the MWL assessment, the picture n-back experiments stimulated by multiple emotions, including negative, neutral and positive emotions, are conducted. The experiment data of 16 participants are collected by the means of task performance, subjective rating and fNIRS physiological measurement. The results show that the MWL of participants is affected by external factors in terms of task performance and subjective rating. An MWL assessment model is established, with a total of 380 physiological features of time domain, frequency domain and nonlinearity domain extracted from the fNIRS information as input and the support vector machine as classifier. Average accuracy of classification is 92.49% for training data which is based on task data stimulated by neutral emotion, and 75.9% and 79.99% for test data which is based on experiment data stimulated by positive and negative emotions. By analyzing the data, the experiments verify the hypothesis that the operator′s MWL is effectively affected by task load and emotional stimulus, demonstrate the feasibility of MWL assessment in different emotional state based on fNIRS, and lay a foundation for the OFS assessment on the strength of the MWL evaluation during complex tasks.