A novel adaptive recovery method in the emerging compressed sensing theory is described and applied to extracellular neural recordings in order to reduce data rate in wireless neural recording systems. To strike a balance between high compression ratio and high spike reconstruction quality, a novel method that employs a group-sparsity recovery algorithm, prior information about the input neural signal, learning prior supports of spikes, and a matched wavelet technique is introduced. Our simulation results, using four different sets of real extracellular recordings from four distinct neural sources, show that our proposed method is effective, viable, and outperforms the state-of-the-art compressed sensing-based methods, in particular, when the number of the measurement is two times of the sparsity.
JOURNAL OF NEUROLOGY AND NEUROSCIENCE ISSN 2171-6625This article is available in: www.jneuro.com of the CS allows the circuit-complexity on the implant side to be moved to the recovery side, which is outside of the body.Furthermore, in spite of the immaturity of the field, it has been shown that energy efficiency, circuit size, and power consumption of a CS encoder are on par with or better than the state-of-the-art of existing compression methods [7][8][9][10][11][12][13][14][15][16].This paper deals with an adaptive recovery method. This part of a CS system takes place outside of the body. Here, to strike a balance between high compression ratio and high spike reconstruction quality, our proposed method is characterized by a number of unique features: 1) the employment of group-sparsity recovery algorithm, 2) taking advantage of prior information about the input signal, 3) the learning of prior supports of spikes, and 4) a matched wavelet technique.An overview and comparison of some existing compression methods are given in Aghagolzadeh and Charbiwala [1,6]. However, in this work, in order to further evaluate the performance of our proposed method, we compare it with two recent works in its category (i.e., techniques employing the CS approach).We recall the basics of the CS in Section II. Section III details our adaptive recovery method. The methodology is presented in Section IV. Then in Section V, our results are reported. An example of a small size and low power cost CS encoder is evinced in Section VI. Finally, we conclude with a discussion of our contribution and perspectives in Section VII.
Compressed Sensing
A. Background of compressed sensingCompressed sensing is a new approach for signal compression. It has been shown that if a signal has a sparse representation in one basis ø , then it can be recovered from a small number of projections onto a second basis φ that is incoherent with the former [17][18][19]. CS is a non-adaptive data acquisition technique because the basis φ does not depend on the measured signal x.Also, to be efficient, CS requires two conditions:Sparse representation: The sparse property of the signal x is very important and directly influences performance of the CS [17][18][19]. G...