The emerging neural-silicon interface devices bridge nerve systems with artificial systems and play a key role in neuro-prostheses and neuro-rehabilitation applications. Integrating neural signal collection, processing and transmission on a single device will make clinical applications more practical and feasible. This paper focuses on the wireless antenna part and real-time neural signal analysis part of implantable brain-machine interface (BMI) devices. We propose to use millimeter-wave for wireless connections between different areas of a brain. Various antenna, including microstrip patch, monopole antenna and substrate integrated waveguide antenna are considered for the intra-cortical proximity communication. A Hebbian eigenfilter based method is proposed for multi-channel neuronal spike sorting. Folding and parallel design techniques are employed to explore various structures and make a trade-off between area and power consumption. Field programmable logic arrays (FPGAs) are used to evaluate various structures.