Abstract - The long-term goal of this research is to improve the functionality and user integration of prostheses, thereby reducing abandonment rates and enhancing user satisfaction. Millions of individuals in the United States live with upper limb loss and turn to prosthetic devices to regain function. However, many abandon their prostheses due to limited functionality and a lack of intuitive control. Neural interfaces that decode signals from residual nerves offer a promising solution by enabling more natural and precise control of advanced prosthetic devices. Our study demonstrates that accurate kinematic predictions can be achieved using only a small subset of neural channels, coupled with straightforward post-processing techniques. By comparing three distinct feature extraction methods, we show that isolating and smoothing neural firing rates leads to enhanced model training and improved prediction accuracy of intended movements. These findings could lead to more effective and accessible prosthetic control systems, lowering the cost and complexity of neural interfaces. Additionally, the techniques developed here can be applied to other fields, such as brain-computer interfaces and rehabilitation devices, to optimize neural decoding and improve patient outcomes.