多通道脑电信号的压缩感知技术研究

发布时间：2018-11-27 15:34

【摘要】：随着微电子、生物医学、脑机接口等领域的发展,以无线体域网为中心,对目标对象进行远程生理信号监测,包括脑电信号、心电信号等逐渐成为常态。而在无线体域网中对多通道脑电信号的长时间监测会受到数据传输节点的带宽、系统功耗的限制。传统数据压缩方法计算复杂,难以在计算能力薄弱的传感节点上实现。压缩感知理论前端压缩编码计算简单,复杂计算主要集中在后端重构,适合于前端多传感节点计算能力弱且对系统功耗和实时性要求高的场景。因此研究多通道脑电信号压缩感知技术具有重要意义。本文的研究目的是在分析压缩感知理论的基础上,结合脑电信号的特点,设计基于FPGA的脑电信号压缩采集系统,实现对多通道脑电信号的实时采集压缩,以解决目前脑电信号压缩感知方法中存在的重构精度低、硬件实现占用存储空间大、通道扩展性差等问题。首先,研究多通道脑电信号的压缩感知方法,分析压缩感知原理和特性,分析脑电信号的稀疏性以及设计模拟随机解调AIC模型和数字测量矩阵模型,分析验证OMP及SOMP重构算法。其次,设计实现基于FPGA的压缩采集系统,硬件采集模块采用ADS1299脑电采集电路,固件设计包括采集、存储、压缩、传输等单元,实现对多通道脑电信号的实时采集压缩功能。再次,设计实现多通道脑电信号重构软件,采用Lab VIEW与MATLAB混合编程的方法,重构软件包括数据的接收和重构两大功能模块,可以实现实时的读取FPGA压缩数据,上位机重构信号的功能。最后,完成了对系统各功能模块的验证和调试,评价脑电压缩系统的性能和精度,并设计实验探讨数据压缩对情感识别率的影响。实验测试结果表明,本文设计的多通道脑电信号压缩感知系统可以实现对脑电信号的实时压缩采集,2倍压缩率SOMP重构平均SNDR为21.74d B,对于脑电情感分类,原始脑电信号情感识别平均分类结果为52.2%,2倍压缩重构数据平均分类结果52.1%,2倍压缩重构对分类结果影响不大。
[Abstract]:With the development of microelectronics, biomedicine and brain-computer interface, remote monitoring of physiological signals, including EEG and ECG signals, has gradually become the norm with the wireless body area network as the center. The long time monitoring of multichannel EEG in wireless body area network is limited by the bandwidth of data transmission node and the power consumption of the system. The traditional data compression method is complex and difficult to be implemented on sensor nodes with weak computing power. Compression sensing theory in front compression coding calculation is simple, complex computing is mainly focused on the back-end reconstruction, suitable for the front-end multi-sensor node computing capacity is weak and the system power consumption and real-time requirements are high. Therefore, it is of great significance to study the multi-channel EEG compression sensing technology. The purpose of this paper is to design an EEG compression acquisition system based on FPGA on the basis of analyzing the theory of compression perception and combining with the characteristics of EEG, so as to realize the real-time acquisition and compression of multichannel EEG signals. In order to solve the problems of low reconstruction precision, large storage space and poor channel expansibility in the current EEG compression and sensing methods, etc. Firstly, the compression sensing method of multichannel EEG signal is studied, the principle and characteristics of compression sensing are analyzed, the sparsity of EEG signal is analyzed, the AIC model of analog random demodulation and the digital measurement matrix model are designed, and the algorithms of OMP and SOMP reconstruction are analyzed and verified. Secondly, the compression acquisition system based on FPGA is designed and implemented. The hardware acquisition module uses ADS1299 EEG acquisition circuit, and the firmware design includes acquisition, storage, compression, transmission and other units to realize the real-time acquisition and compression of multi-channel EEG signals. Thirdly, the multi-channel EEG signal reconstruction software is designed and implemented. The software includes two function modules of data receiving and reconstructing, which can read FPGA compressed data in real time by using the mixed programming method of Lab VIEW and MATLAB. The function of the upper computer to reconstruct the signal. Finally, the function modules of the system are verified and debugged, the performance and accuracy of EEG compression system are evaluated, and the influence of data compression on emotion recognition rate is discussed by designing experiments. The experimental results show that the multi-channel EEG compression sensing system designed in this paper can realize the real-time compression and acquisition of EEG signals. The average SNDR of SOMP reconstruction with 2 times compression ratio is 21.74 dB. The average classification result of original EEG emotion recognition is 52.2 times compression reconstruction data average classification result 52.1 times compression reconstruction has little effect on classification result.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R318;TN911.7

【相似文献】