Ten million people on the earth at any given time suffer from the loss or lack of a limb. Three million of these are upper extremity amputees. No matter the cause or reason for amputation the amputees’ life is never the same again. Independence can be lost and never regained. The loss of a limb can be an overpowering event that effects the mental, social, and physical well-being of a person. Although there has been many attempts to restore hand function through neural prosthetics there is little to no credible solutions to regain it. This thesis presents methods to restore hand function based on the human grasping system and addresses key challenges in the practical application of those methods. The contribution of this work lies in improving artificial hand function by mimicking the human grasping system. The human grasping system consists of the brain, the nervous system and the hand. The methods used investigate: biomechanical design principles for artificial hands, the limitations of machine learning in non-invasive neural interfaces (NI’s) and the development of an autonomous hand control framework for artificial hand control. The application of the Support Vector Machine (SVM) to a non-invasive neural interface is principally limited to binary applications in the real world. In order to overcome this limitation an autonomous approach is developed to achieve better response and accuracy from NI control frameworks. In binary applications the SVM performed well with a classification accuracy of around 80% accompanied by a response time of just under 5 seconds. When applied to multiple classifications the SVM decreased in accuracy and response dramatically. The accuracy of the autonomous hand control framework had a classification accuracy of 95% with a response time comparable to that of the SVM. This outcome has the potential to move non-invasive NI research into the real world and to restore hand function in an intuitive and meaningful way. A key learning of this research is that machine learning methods for non-invasive neural interfaces are not the best way to restore hand function for amputees. Hand function is best restored through the combining of machine learning and machine vision into a single NI based control framework. Additionally, this work is based in Mātauranga Māori with a western science approach. By so doing, amputees will be empowered, have increased independence and enjoy a better quality of life. The scope of this work is limited to the consideration of hand function restoration for upper extremity amputees. In particular trans-radial amputees. Focus is given to non-invasive methods of control, therefore, investigations are limited to the electroencephalography (EEG) signal. Mimicking the human grasping system is a complex and difficult process, therefore, the attempts in this work to restore hand function only consider mechanisms and processes related directly to the motor control of the human hand.