In order to improve the online recognition accuracy of the grinding direction of single crystal diamond tools and address the limitation of acquiring limited information from a single sensor in grinding monitoring, this study proposes a method for online recognition of the grinding direction of single crystal diamond tools based on multi-information fusion and particle swarm optimization (PSO) algorithm for optimizing the BP neural network. Vibration signals and acoustic emission (AE) signals were collected during the grinding process. The wavelet packet decomposition method was applied to analyze the vibration signals of the tool and identify the characteristic frequency bands strongly correlated with the grinding direction. The parameter analysis method was used to analyze the AE signals and extract the characteristic parameters. The energy values of the characteristic frequency bands in the vibration signals and the characteristic parameters of the AE signals were taken as the feature parameters for identifying the grinding direction of the tool. These feature parameters were then used as inputs to the BP neural network model for fusion and online recognition of the grinding direction. To overcome the disadvantage of the BP neural network easily getting stuck in local minima, the PSO algorithm was utilized to optimize the weights and thresholds of the neural network, effectively solving the problem of local minima. Experimental results demonstrated that the PSO-BP neural network combined with multi-information fusion significantly improved the online recognition accuracy of the grinding direction of single crystal diamond tools, achieving an accuracy of 85%. This proposed method provides a new approach for online recognition of the grinding direction of single crystal diamond tools.