A reduced complexity integer lifting wavelet-based module for real-time processing in implantable neural interface devices

In high-density implantable microelectrode arrays, the sheer amount of data recorded poses a serious challenge to the bandwidth of the associated biotelemetry system. The power limitations and restricted chip area for an implantable module strongly imposes significant limitations on the computational complexity. In this work, we propose a new methodology for processing multichannel neural data most suitable for hardware implementation. Our work builds on previous work aimed at using the discrete wavelet transform (DWT) compact representation of the acquired neural data. We propose a novel design of the DWT computational block in a neural interface module and evaluate its performance in terms of speed, complexity and fidelity of the reconstructed signals. We show that most of the computations in this block can be performed with fixed-point representations and finite precision without significant loss in performance compared to the traditional floating-point computations in standard DSP chips.