Yuanming Suo

An Efficient and Compact Compressed Sensing Microsystem for Implantable Neural Recordings

Multi-Electrode Arrays (MEA) have been widely used in neuroscience experiments. However, the reduction of their wireless transmission power consumption remains a major challenge. To resolve this challenge, an efficient on-chip signal compression method is essential. In this paper, we first introduce a signal-dependent Compressed Sensing (CS) approach that outperforms previous works in terms of compression rate and reconstruction quality. Using a publicly available database, our simulation results show that the proposed system is able to achieve a signal compression rate of 8 to 16 while guaranteeing almost perfect spike classification rate. Finally, we demonstrate power consumption measurements and area estimation of a test structure implemented using TSMC 0.18 μm process. We estimate the proposed system would occupy an area of around 200 μm ×300 μm per recording channel, and consumes 0.27 μW operating at 20 KHz .

Structured Sparse Priors for Image Classification

Model-based compressive sensing (CS) exploits the structure inherent in sparse signals for the design of better signal recovery algorithms. This information about structure is often captured in the form of a prior on the sparse coefficients, with the Laplacian being the most common such choice (leading to l1-norm minimization). Recent work has exploited the discriminative capability of sparse representations for image classification by employing class-specific dictionaries in the CS framework. Our contribution is a logical extension of these ideas into structured sparsity for classification. We introduce the notion of discriminative class-specific priors in conjunction with class specific dictionaries, specifically the spike-and-slab prior widely applied in Bayesian sparse regression. Significantly, the proposed framework takes the burden off the demand for abundant training image samples necessary for the success of sparsity-based classification schemes. We demonstrate this practical benefit of our approach in important applications, such as face recognition and object categorization.

Multi-task image classification via collaborative, hierarchical spike-and-slab priors

Promising results have been achieved in image classification problems by exploiting the discriminative power of sparse representations for classification (SRC). Recently, it has been shown that the use of class-specific spike-and-slab priors in conjunction with the class-specific dictionaries from SRC is particularly effective in low training scenarios. As a logical extension, we build on this framework for multitask scenarios, wherein multiple representations of the same physical phenomena are available. We experimentally demonstrate the benefits of mining joint information from different camera views for multi-view face recognition.

Energy-Efficient Multi-Mode Compressed Sensing System for Implantable Neural Recordings

Widely utilized in the field of Neuroscience, implantable neural recording devices could capture neuron activities with an acquisition rate on the order of megabytes per second. In order to efficiently transmit neural signals through wireless channels, these devices require compression methods that reduce power consumption. Although recent Compressed Sensing (CS) approaches have successfully demonstrated their power, their full potential is yet to be explored. Built upon our previous on-chip CS implementation, we propose an energy efficient multi-mode CS framework that focuses on improving the off-chip components, including (i) a two-stage sensing strategy, (ii) a sparsifying dictionary directly using data, (iii) enhanced compression performance from Full Signal CS mode and Spike Restoration mode to Spike CS + Restoration mode and; (iv) extension of our framework to the Tetrode CS recovery using joint sparsity. This new framework achieves energy efficiency, implementation simplicity and system flexibility simultaneously. Extensive experiments are performed on simulation and real datasets. For our Spike CS + Restoration mode, we achieve a compression ratio of 6% with a reconstruction SNDR > 10 dB and a classification accuracy > 95% for synthetic datasets. For real datasets, we get a 10% compression ratio with ~ 10 dB for Spike CS + Restoration mode.

A closed-loop compressive-sensing-based neural recording system

OBJECTIVE This paper describes a low power closed-loop compressive sensing (CS) based neural recording system. This system provides an efficient method to reduce data transmission bandwidth for implantable neural recording devices. By doing so, this technique reduces a majority of system power consumption which is dissipated at data readout interface. The design of the system is scalable and is a viable option for large scale integration of electrodes or recording sites onto a single device. APPROACH The entire system consists of an application-specific integrated circuit (ASIC) with 4 recording readout channels with CS circuits, a real time off-chip CS recovery block and a recovery quality evaluation block that provides a closed feedback to adaptively adjust compression rate. Since CS performance is strongly signal dependent, the ASIC has been tested in vivo and with standard public neural databases. MAIN RESULTS Implemented using efficient digital circuit, this system is able to achieve >10 times data compression on the entire neural spike band (500-6KHz) while consuming only 0.83uW (0.53 V voltage supply) additional digital power per electrode. When only the spikes are desired, the system is able to further compress the detected spikes by around 16 times. Unlike other similar systems, the characteristic spikes and inter-spike data can both be recovered which guarantes a >95% spike classification success rate. The compression circuit occupied 0.11mm(2)/electrode in a 180nm CMOS process. The complete signal processing circuit consumes <16uW/electrode. SIGNIFICANCE Power and area efficiency demonstrated by the system make it an ideal candidate for integration into large recording arrays containing thousands of electrode. Closed-loop recording and reconstruction performance evaluation further improves the robustness of the compression method, thus making the system more practical for long term recording.

Hierarchical sparse modeling using Spike and Slab priors

Sparse modeling has demonstrated its superior performances in many applications. Compared to optimization based approaches, Bayesian sparse modeling generally provides a more sparse result with a knowledge of confidence. Using the Spike and Slab priors, we propose the hierarchical sparse models for the scenario of single task and multitask - Hi-BCS and CHi-BCS. We draw the connections of these two methods to their optimization based counterparts and use expectation propagation for inference. The experiment results using synthetic and real data demonstrate that the performance of Hi-BCS and Chi-BCS are comparable or better than their optimization based counterparts.

Group structured dirty dictionary learning for classification

Dictionary learning techniques have gained tremendous success in many classification problems. Inspired by the dirty model for multi-task regression problems, we proposed a novel method called group-structured dirty dictionary learning (GDDL) that incorporates the group structure (for each task) with the dirty model (across tasks) in the dictionary training process. Its benefits are two-fold: 1) the group structure enforces implicitly the label consistency needed between dictionary atoms and training data for classification; and 2) for each class, the dirty model separates the sparse coefficients into ones with shared support and unique support, with the first set being more discriminative. We use proximal operators and block coordinate decent to solve the optimization problem. GDDL has been shown to give state-of-art result on both synthetic simulation and two face recognition datasets.

Real Time Compressive Sensing Video Reconstruction in Hardware

Compressive sensing has allowed for reconstruction of missing pixels in incomplete images with higher accuracy than was previously possible. Moreover, video data or sequences of images contain even more correlation, leading to a much sparser representation as demonstrated repeatedly in numerous digital video formats and international standards. Compressive sensing has inspired the design of a number of imagers which take advantage of the need to only subsample a scene, which reduces power consumption by requiring acquisition and transmission of fewer samples. In this paper, we show how missing pixels in a video sequence can be estimated using compressive sensing techniques. We present a real time implementation of our algorithm and show its application to an asynchronous time-based image sensor (ATIS) from the Austrian Institute of Technology. The ATIS only provides pixel intensity data when and where a change in pixel intensity is detected, however, noise randomly causes intensity changes to be falsely detected, thereby providing random samples of static regions of the scene. Unlike other compressive sensing imagers, which typically have pseudo-random sampling designed in at extra effort, the ATIS used here provides random samples as a side effect of circuit noise. Here, we describe and analyze a field-programmable gate array implementation of a matching pursuit (MP) algorithm for compressive sensing reconstruction capable of reconstructing over 1.9 million 8 × 8 pixel regions per second with a sparsity of 11 using a basis dictionary containing 64 elements. In our application to ATIS we achieve throughput of 28 frames per second at a resolution of 304 × 240 pixels with reconstruction accuracy comparable to that of state of the art algorithms evaluated offline.

A dictionary learning algorithm for multi-channel neural recordings

Multi-channel neural recording devices are widely used for in vivo neuroscience experiments. Incurred by high signal frequency and large channel numbers, the acquisition rate could be on the order of hundred MB/s, which requires compression before wireless transmission. In this paper, we adopt the Compressed Sensing framework with a simple on-chip implementation. To improve the performance while reducing the number of measurements, we propose a multi-modal structured dictionary learning algorithm that enforces both group sparsity and joint sparsity to learn sparsifying dictionaries for all channels simultaneously. When the data is compressed 50 times, our method can achieve a gain of 4 dB and 10 percentage units over state-of-art approaches in terms of the reconstruction quality and classification accuracy, respectively.

Energy-efficient two-stage Compressed Sensing method for implantable neural recordings

For in-vivo neuroscience experiments, implantable neural recording devices have been widely used to capture neural activity. With high acquisition rate, these devices require efficient on-chip compression methods to reduce power consumption for the subsequent wireless transmission. Recently, Compressed Sensing (CS) approaches have shown great potentials, but there exists the tradeoff between the complexity of the sensing circuit and its compression performance. To address this challenge, we proposed a two-stage CS method, including an on-chip sensing using random Bernoulli Matrix S and an off-chip sensing using Puffer transformation P. Our approach allows a simple circuit design and improves the reconstruction performance with the off-chip sensing. Moreover, we proposed to use measureed data as the sparsifying dictionary D. It delivers comparable reconstruction performance to the signal dependent dictionary and outperforms the standard basis. It also allows both D and P to be updated incrementally with reduced complexity. Experiments on simulation and real datasets show that the proposed approach can yield an average SNDR gain of more than 2 dB over other CS approaches.