應用於腦電圖無失真壓縮混合訊號積體電路設計

Mixed-Signal Integrated Circuit Design for Lossless EEG Compression

蔡孟潔(MENG-CHIEH TSAI)

指導教授 ： 陳世綸

中原大學/電機資訊學院/電子工程學系/碩士(2023年)

https://doi.org/10.6840/cycu202301098

摘要

為了因應更多元的使用需求，近年來無線的穿戴式裝置快速發展，而在量測生理訊號時，大多數的受測者在乎的是訊號收集時的舒適度及便利性。然而監測生理訊號裝置的體積大小主要因為裝置傳輸時的功率消耗而受限於其電池的體積，為了要更進一步縮小監測裝置的體積，勢必需要將資料壓縮在進行傳輸以減少功率消耗。
為了達到以上所述之目的，我們利用 TSMC 0.18-μm CMOS Mixed Signal 製程完成一個10位元的逐漸逼近式類比數位轉換器晶片，總面積為 963.1×544.6 〖μm〗^2 ，取樣率為1MHz。而在壓縮訊號演算法的部分，本論文提出一種以電路實現為目的之以閥值評等訊號的趨勢等級結合動態哥倫布編碼的低複雜度無失真腦電圖壓縮演算法。本論文選擇使用CHB-MIT Scalp EEG Database中的腦波訊號作為資料來源，對提出的演算法進行評估與比較。考量腦電圖演算法積體電路的實現，及使壓縮技術達到即時且無失真的效果，演算法以低複雜度的方式設計，降低其壓縮演算法的運算量，進而減少電路實現的成本及維持電路的效能。
利用不同閥值定義訊號的趨勢等級作為預測演算法的核心，並以投票預測來選擇最佳的預測方程式，決定如何從先前的數據中預測當前數據，再加上校正作用的最佳預測常數Correction parameter C，來達到最佳預測效果。接著使用動態哥倫布萊斯編碼重新將預測結果編碼成可變長度碼，此方法能有效提升資料壓縮效果。與先前的文獻相比，本論文提出的壓縮演算法能在23個channels中使腦電圖無失真平均壓縮率達到2.34。
在未來可以將這個壓縮系統應用在穿戴式腦波儀，縮小其體積，使這類裝置在日常生活中能更加普及。

關鍵字

腦電圖 ； 逐漸逼近式類比數位轉換器 ； 無失真 ； 閥值評等 ； 哥倫布萊斯編碼

並列摘要

In recent years, wireless wearable devices have developed rapidly in order to meet more diverse usage requirements. When measuring physiological signals, most of the subjects care about the comfort and convenience of signal collection. However, the size of the physiological signal monitoring device is mainly limited by the size of its battery due to the power consumption of the device during transmission. In order to further reduce the size of the device, it is necessary to compress the data for transmission to reduce power consumption.
In order to achieve this goal, in this thesis we use TSMC 0.18-μm CMOS Mixed Signal process to complete a 10-bit SAR ADC chip with a total area of 963.1×544.6 〖μm〗^2 and the sampling rate is 1MHz. This thesis proposed a lossless EEG compression algorithm with threshold leveling and dynamic Golomb-Rice coding. The CHB-MIT Scalp EEG Database was selected as testing dataset in this thesis for evaluating the algorithm performance. Considering integrated circuit implementation, the proposed algorithm is realized with low complex operation for performance and low-cost of circuit. In this way, the operation of the compression algorithm is reduced, thereby reducing the cost of implementing the circuit and maintaining the performance of the circuit.
In prediction, the thresholds define and level the signal. These levels efficiently perform the trend of EEG signal. The best prediction equation and correction parameter are selected by voting prediction to achieve the best compression effect. With this correction parameter C, the prediction method greatly improves accuracy. For presenting predicted data, dynamic Golomb-Rice coding generates a variable length binary code for the result of prediction. Combining these methods, the algorithm predicts the signal trend in efficient and greatly improves the compression rate of original data. Compared with previous works, the results show that the proposed algorithm achieves average compression rate 2.34 of 23 channels.
In the future, it is hoped that this compression system can be applied to wearable EEG devices to reduce their size and make such devices more popular in daily life.

並列關鍵字

EEG ； SAR ADC ； lossless ； threshold leveling ； Golomb-Rice coding

參考文獻

[1]. Liu, C. C., Chang, S. J., Huang, G. Y., & Lin, Y. Z. (2009, June). A 0.92 mW 10-bit 50-MS/s SAR ADC in 0.13 μm CMOS process. In 2009 symposium on VLSI circuits (pp. 236-237). IEEE.

[2]. Cho, Y. K., Jeon, Y. D., Nam, J. W., & Kwon, J. K. (2010). A 9-bit 80 MS/s successive approximation register analog-to-digital converter with a capacitor reduction technique. IEEE Transactions on Circuits and Systems II: Express Briefs, 57(7), 502-506.

[3]. Wen, T., & Zhang, Z. (2018). Deep convolution neural network and autoencoders-based unsupervised feature learning of EEG signals. IEEE Access, 6, 25399-25410.

[4]. Chen, C. A., Chen, S. L., Huang, H. Y., & Luo, C. H. (2012). An efficient micro control unit with a reconfigurable filter design for wireless body sensor networks (WBSNs). sensors, 12(12), 16211-16227.

[5]. Chen, S. L., Luo, G. A., & Lin, T. L. (2013). Efficient fuzzy‐controlled and hybrid entropy coding strategy lossless ECG encoder VLSI design for wireless body sensor networks. Electronics Letters, 49(17), 1058-1060.