使用802.15.4無線感測網路建立多通道神經信號記錄系統

Abstract

腦神經科學家使用神經信號記錄系統對數百個神經元做神經信號記錄，藉由分析信號，了解複雜的神經網路架構與彼此之間的關係。在對清醒且具自由行為能力動物做神經信號記錄的過程中，植入在動物大腦的微電極與記錄系統之間信號經由連接線傳遞。然而，有線神經信號記錄系統架構下，因連接線的拉扯，必然使動物體的行為受到限制。除此之外，連接線也侷限了實驗操作的環境。因此，結合射頻技術發展出的無線神經信號記錄系統是較為適合用於對清醒且具自由行為能力動物的神經信號做記錄。近年來，多動物體社交行為時與其大腦的神經行為之間的關係被受科學家關注。本研究致力於發展星狀拓樸架構下的低資料量無線感測網路系統用於多生物體之無線神經信號記錄。 在系統設計之中，引用適應性門檻值方法，期望降低神經信號中雜訊振幅突然增加時，導致錯誤神經動作電位偵測的情形；並將此演算法實作在FPGA中，處理多通道神經信號。被系統偵測得的神經動作電位波形與其相對應發生時間和發生的通道等資訊，組裝成封包傳送至用戶端顯示與儲存。考慮星狀拓樸中，多個資料傳送節點同時傳送資料，造成封包碰撞，導致資料遺失的情形，在無線封包傳輸機制上，加入時間排程機制(Transmiision Time Schedule, TTS)，期望降低資料遺失的機率。 本研究利用機率與期望值分析星狀拓樸網路中的有效頻寬與傳輸封包時間，進而推導出最適合TTS架構的一個時槽時間。同時分析自由競爭傳輸機制(Carrier Sense Multiple Access with Collision Avoidance, CSMA/CA)與TTS架構之間的效能與傳輸成功率。在節點對節點拓樸架構下，針對不同的FPGA資料處理(多通道神經動作電位偵測)後的資料量輸出率對Spike latency的影響做模擬。在星狀拓樸架構中使用CSMA/CA與TTS二個傳送機制，實測不同的FPGA資料處理後資料量輸出率對Spike latency與封包遺失率的影響，再比對模擬與實測結果，結論出不同拓樸架構與不同傳送機制時，系統所能負荷的FPGA資料輸出率範圍。值得一提的是，多通道神經信號經神經動作電位偵測處理後，資料量的下降比率為15倍之多，大大的降低無線網路資料傳輸的負擔，進而提升了系統的效能。而在星狀拓樸多資料傳送節點架構下，由於TTS傳輸機制的引入，使得資料傳送成功率達到將近100%。

With the utilization of neural recording system, recording from hundreds of neurons will allow neurophysiologists to better understand the complex circuits of the nerve system. Contemporary experimental neural recording systems aimed for a free-moving animal connect implanted microelectrodes to external processing equipment through a cable tether. On the wired paradigm, the animal must behaviorally contented with the sensation of being tethered. Besides, the long cables also limit the size of the recording environment. It is therefore clear that replacing the wired instruments with telemetry interface is better suited to record neural signals from the free-moving animal for chronic recording. In recent years, the relationship between the social behavior of animal and the neural activities has been attractive to many neurophysiologists in the world. In this study, neural recording system based on wireless sensor network where the star topology is adopted is designed for processing and recording the multi-channel neural signals of multi-subjects.The adaptive-threshold spike-detection method is included to produce the beter performance of spike-detection as the burst on the scale of noise exists. The waveform and timing of spikes resulted from hardware-implemented adaptive-threshold spike-detection using FPGA is collected and wirelessly transmitted to Client. Considering the packet collision in the start topology, the transmission time scheduling (TTS) protocol is served as scheduling the timing of transmitted packets of each node in the network. In result, this study utilizes simulations and experiments to analysis the effective bandwidth, spike latency with different number of channel, and comparing the performance of the proposed neural recording system with TTS scheme and without TTS scheme. In conclusion, the ratio of data reduction performed by spike-detection algorithm is about 15. The limitation of FPGA output data rate resulted from the multi-channl spike-detection method is been found out depending on the scale of star topology with TTS and without TTS. With the usage of TTS scheme, the packet loss rate is about zero whatever the number of data transmitter node in the star topology is.