Artificial intelligence is an essential demand in several domains, including cryptographic security systems, video attendance, object recognition and tracking systems, etc. Multiplication is the operation that takes up a lot of execution time when processing an AI model, so it is crucial to optimize the model to calculate quickly in order to reduce power consumption while maintaining appropriate performance when deploying AI models on edge devices with limited computing resources and power sources. Due to the promising efficiency of the approximate multiplier circuit, there are several research on novel approximate multiplication models that have been proposed and implemented on edge devices in general and FPGA-based Edge devices in particular. In this study, the approximation multiplier approaches for FPGA systems is systematically evaluated. The system compares the energy usage, latency, and area of different methods and also present future research directions for applying approximate multiplier techniques for Edge AI.