Yuning Yang

Low Power Multimode Electrochemical Gas Sensor Array System for Wearable Health and Safety Monitoring

This paper presents an electrochemical gas sensor array system for health and safety monitoring. The system incorporates a custom room temperature ionic-liquid gas sensor array, a custom multimode electrochemical sensor readout board, and a commercial low power microcontroller board. Sensors for multiple gas targets were implemented in a miniaturized 2 × 2 array where each sensor consumes <;3.2 μW and occupies a sensing area volume of 350 mm3. A novel resource-sharing circuit architecture tailored to the gas sensor array was utilized to significantly decrease power, cost, and size. The system supports multiple electrochemical measurement modes to provide orthogonal data to in-module sensor array algorithms for better prediction accuracy. The system achieves a resolution as high as 0.01 vol% in amperometry mode and 0.06 vol% in ac impedance mode for oxygen as an example target gas.

Adaptive threshold spike detection using stationary wavelet transform for neural recording implants

Spike detection is an essential first step in the analysis of neural recording signals. A new spike detection hardware architecture combining absolute threshold method and stationary wavelet transform (SWT) is described. The method enables spike detection with 90% accuracy even when the signal-to-noise is −1dB. A noise monitoring block was implemented to automatically calculate the appropriate threshold value for spike detection, and the system then chooses either absolute threshold method or the SWT method to optimize power consumption. The system was designed in 130nm CMOS and shown to occupy 0.082 mm2 and dissipate 0.45 μW for one channel.

Adaptive Threshold Neural Spike Detector Using Stationary Wavelet Transform in CMOS

Spike detection is an essential first step in the analysis of neural recordings. Detection at the frontend eases the bandwidth requirement for wireless data transfer of multichannel recordings to extra-cranial processing units. In this work, a low power digital integrated spike detector based on the lifting stationary wavelet transform is presented and developed. By monitoring the standard deviation of wavelet coefficients, the proposed detector can adaptively set a threshold value online for each channel independently without requiring user intervention. A prototype 16-channel spike detector was designed and tested in an FPGA. The method enables spike detection with nearly 90% accuracy even when the signal-to-noise ratio is as low as 2. The design was mapped to 130 nm CMOS technology and shown to occupy 0.014 mm2 of area and dissipate 1.7 μW of power per channel, making it suitable for implantable multichannel neural recording systems.

Power-error analysis of sensor array regression algorithms for gas mixture quantification in low-power microsystems

Reliable gas sensors are highly desired for many applications, but their typically poor specificity requires arrays of cross-sensitive sensors to predict identity and concentrations of gas mixtures. A relationship between sensor outputs and gas concentrations can be formulated using regression models. This paper presents a detailed analysis of regression models generated using different algorithms. The analysis incorporates a variety of sensor parameters as well as the power consumption of each model when implemented within a low-power microcontroller. The results provide new insight into the effects of sensor array parameters on prediction errors and the tradeoffs between prediction errors and power for different regression models.

Power-area efficient VLSI implementation of decision tree based spike classification for neural recording implants

Spike classification is the last step in spike sorting to reduce the data rate of a brain-machine interface. This paper presents a new decision tree based spike classification method that achieves a classification accuracy comparable to methods based on L1 distance. The design was synthesized for 130nm CMOS with an architecture that interleaves eight channels to optimize the power-area tradeoff. Resource analysis shows that the resulting design consumes 32nW of power per channel at a clock rate of 50KHz and occupies 5115μm2 of area per channel.

Design of a configurable neural Data compression system for intra-cortical implants

Multi-channel neural signal recordings need high data compression and efficient data transmission. Our previous work has shown a practical data compression solution based on discrete wavelet transform, multi-level thresholding and run length encoding. This paper presents a custom designed communication protocol for bidirectional data telemetry to and from the implanted module. A global controller is also presented which configures, operates and unites all the modules together effectively and efficiently into a 32-channel system. Performance of the communication protocol and the compression engine is analyzed.

A configurable realtime DWT-based neural data compression and communication VLSI system for wireless implants

This paper presents the design of a complete multi-channel neural recording compression and communication system for wireless implants that addresses the challenging simultaneous requirements for low power, high bandwidth and error-free communication. The compression engine implements discrete wavelet transform (DWT) and run length encoding schemes and offers a practical data compression solution that faithfully preserves neural information. The communication engine encodes data and commands separately into custom-designed packet structures utilizing a protocol capable of error handling. VLSI hardware implementation of these functions, within the design constraints of a 32-channel neural compression implant, is presented. Designed in 0.13μm CMOS, the core of the neural compression and communication chip occupies only 1.21mm(2) and consumes 800μW of power (25μW per channel at 26KS/s) demonstrating an effective solution for intra-cortical neural interfaces.

Implantable neural spike detection using lifting-based stationary wavelet transform

Spike detection from high data rate neural recordings is desired to ease the bandwidth bottleneck of bio-telemetry. An appropriate spike detection method should be able to detect spikes under low signal-to-noise ratio (SNR) while meeting the power and area constraints of implantation. This paper introduces a spike detection system utilizing lifting-based stationary wavelet transform (SWT) that decomposes neural signals into 2 levels using ‘symmlet2’ wavelet basis. This approach enables accurate spike detection down to an SNR of only 2. The lifting-based SWT architecture permits a hardware implementation consuming only 6.6 μW power and 0.07mm2 area for 32 channels with 3.2 MHz master clock.

Identification and quantification of mixed air pollutants based on homotopy method for gas sensor array

Accurate recognition of air pollutants and estimation of their concentrations are critical for human health and safety monitoring and can be achieved using gas sensor arrays. In this paper, an efficient method based on a homotopy algorithm is presented for the analysis of sensor arrays responding to binary mixtures. The new method models the responses of a gas sensor array as a system of nonlinear equations and provides a globally convergent way to find the solution of the system. Real data measurement for CH4 and SO2 are used to model sensor responses. The model is applied to the method for prediction and it shows the prediction results are within 1% variation of true values for both gas models.

Frequency Band Separability Feature Extraction Method With Weighted Haar Wavelet Implementation for Implantable Spike Sorting

Hardware-efficient feature extraction is an important step for real-time and on-chip spike sorting. Based on an analysis of spike energy spectrum, a new feature set is developed using the positive and negative spike peaks in low and high frequency bands. A separability metric that evaluates the informativeness and noise sensitivity of features is introduced to optimize the cutoff frequency of each band. Haar-based discrete wavelet transform was chosen to implement memory- and hardware-efficient filters for extracting frequency band separability features. Specifically, peaks from the first level detail and the fourth level approximation were used to represent a spike. To improve clustering performance, the detail features were weighted into the same dynamic range as the approximation features. The new feature extraction method was tested at different signal-to-noise ratios using synthesized datasets consisting of considerable and various spike shapes extracted from real neural recordings. The results show that the new method has 3%–10% better spike sorting performance than other hardware-efficient methods while consuming comparable hardware resources.