Research on Real Time EEG Signal Processing Chip Based on FPGA

Tianle Shi

College of Science, University of Shanghai for Science and Technology, Shanghai, 200093, China

Keywords: FPGA, EEG Signal Processing, Real‑Time Classification.

Abstract: This paper examines state-of-the-art in real-time electroencephalography (EEG) signal processing using Field

Programmable Gate Arrays (FPGAs). EEG is a powerful tool for under-standing brain activity but presents

challenges due to its low amplitude, high noise levels, and complex patterns. Traditional software-based

methods are increasingly inadequate for real-time applications, such as portable brain-computer interfaces

and wireless body area networks, due to computational inefficiencies and power consumption. FPGAs offer

a transformative solution through their hardware-parallel architecture, dynamic reconfigurability, and

computational efficiency. This review highlights key applications of FPGAs in EEG signal preprocessing,

feature extraction, classification, and integrated system design. Comparative analysis shows significant

improvements in processing speed, resource utilization, and classification accuracy over traditional methods.

However, challenges remain in optimizing resource utilization, expanding multimodal applications, and

enhancing portability. Future research directions include developing more efficient algorithms, advanced

multimodal systems, and practical FPGA-based solutions to meet diverse real-world needs. The advancements

discussed in this review are crucial for advancing the field of neurotechnology, as they have the potential to

revolutionize healthcare and human-computer interaction by enabling more efficient and portable EEG-based

systems. By addressing the limitations of traditional methods, this work paves the way for real-time, low-

power, and high-accuracy applications that can significantly improve clinical diagnostics and brain-computer

interface performance.

1 INTRODUCTION

Electroencephalography (EEG) has emerged as a

powerful tool for understanding the complex

dynamics of brain activity, offering a non-invasive

window into the neural processes that underpin

human cognition, emotion, and behavior. By

capturing the electrical signals generated by the

brain’s neurons, EEG provides a high temporal

resolution view of brain function, making it

invaluable for both clinical diagnostics and research

applications. EEG has become an essential

technology in modern neuroscience and healthcare. It

can identify neurological disorders and facilitate

advanced brain–computer interfaces for

communication and control (Shyu et al., 2013).

However, the effective utilization of EEG signals

requires sophisticated processing techniques to

extract meaningful information from the raw data.

EEG signals are often characterized by their low

amplitude, high noise levels, and complex patterns.

To address these challenges, robust preprocessing,

feature extraction, and classification methods are

essential. Traditional approaches, primarily based on

software implementations, have long dominated the

field but are increasingly limited by their

computational inefficiencies, power consumption,

and inability to handle real-time processing demands.

These limitations are particularly pronounced in

emerging applications such as portable BCIs, real-

time neurofeedback systems, and wireless body area

networks (WBANs), where low latency, high

accuracy and energy efficiency are critical (Lin et al.,

2013).

The advent of Field Programmable Gate Arrays

(FPGAs) has introduced a transformative solution to

these challenges. FPGAs are highly parallel,

reconfigurable hardware platforms that offer

unparalleled flexibility and efficiency for real-time

signal processing tasks. Their ability to implement

custom algorithms with both low power consumption

and high computational speed makes them ideal for

processing the intricate and dynamic EEG signals.

Shi, T.

Research on Real Time EEG Signal Processing Chip Based on FPGA.

DOI: 10.5220/0014297500004933

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Biomedical Engineering and Food Science (BEFS 2025), pages 5-9

ISBN: 978-989-758-789-4

Over the past decade, significant advancements in

FPGA technology, including improved processing

capabilities and reduced power consumption, have

led to a surge of research exploring their potential for

EEG signal processing applications. For example, the

CereBridge system, an efficient FPGA-based real-

time processing platform for mobile BCIs, has

demonstrated the feasibility of acquiring and fully

processing up to 32 EEG channels with high precision

and low power consumption (Wahalla et al., 2020).

Another study implemented a seizure detection and

control system using FPGAs, showing significant

improvements in power efficiency and logic element

utilization (Tamilarasi et al., 2018). These studies

have demonstrated that FPGA-based systems can

significantly out-perform traditional software

implementations, showing up to 10x improvements in

processing speed, 30% better resource utilization, and

15% higher classification accuracy (Aravind et al.,

2016).

This review aims to provide a comprehensive

overview of the state-of-the-art in real-time EEG

signal processing using FPGAs. This paper will

examine how FPGAs are being applied across various

stages of EEG signal processing, including

preprocessing, feature extraction, classification, and

integrated system design. By analyzing key studies

and comparing FPGA-based solutions with

traditional methods, this paper will highlight the

unique advantages of FPGA technology in terms of

processing speed, power efficiency, and real-time

performance in this do-main. Additionally, this paper

will discuss the current challenges and future

directions for further ad-vancing FPGA-based EEG

signal processing systems, with a focus on improving

resource utilization, expanding multimodal

applications, and enhancing portability and

practicality for diverse real-world scenarios.

2 RESEARCH ON REAL TIME

EEG SIGNAL PROCESSING

CHIP

2.1 Application of FPGA in EEG

Signal Preprocessing

In EEG signal processing, preprocessing is one of the

key steps, which aims to remove noise and artifacts

and improve signal quality. Kalyana Sundaram et al.

reported an FPGA based EEG signal preprocessing

system, which uses moving average filter and median

filter to re-move noise (Sundaram et al., 2016). The

experimental results show that the median filter

consumes less hardware resources and has lower

power consumption while removing noise.

Specifically, the median filter only occupies about

60% of logical units (LUTs) when processing 256

data points, while the moving average filter occupies

82% of logical units. In addition, as shown in Figure

1, the dynamic power consumption of the median

filter is 0.033 mW, while the dynamic power

consumption of the moving average filter is 0.034

mW. These results indicate that the median filter has

significant advantages in hardware implementation.

Figure 1. Power Vs Temperature for Median filter

(Sundaram et al., 2016).

Besides noise removal, signal separation is another

crucial preprocessing task. Dongsheng Zhao et al.

proposed a Fast Independent Component Analysis

(FastICA) algorithm based on FPGA for separating

mixed EEG signals (Zhao et al., 2015). This

algorithm achieves efficient signal separa-tion

through a macro pipeline architecture, and

experimental results show that it outperforms

traditional FPGA implementations in terms of

processing speed and resource utilization.

Specifically, the algorithm takes 0.0025 seconds to

process EEG signals with 4 channels and a sampling

rate of 10kHz, while traditional FPGA

implementations take 0.003 seconds to process. In

addition, the algorithm also performs well in resource

utilization, occupying only 57% of logical units,

while traditional FPGA implementations occupy 70%

of logical units. These results indicate that the macro

pipeline architecture has significant advantages in

improving processing speed and resource utilization.

Feature extraction is another important step in

EEG signal processing, aimed at extracting useful

features from preprocessed signals for subsequent

BEFS 2025 - International Conference on Biomedical Engineering and Food Science

classification. Xunguang Ma et al. proposed an FPGA

based Convolutional Neural Network (CNN)

accelerator for classifying mo-tor imagery (MI) EEG

signals (Ma et al., 2019). The accelerator implements

a synchronous data stream (SDF) model to realize a

16-bit fixed-point CNN. Experimental results show

that the accelerate-tor outperforms traditional PC

implementations in terms of classification accuracy

and processing speed. Specifically, the accelerator

achieved an average classification accuracy of 80.5%

when processing dataset I in BCI Competition

Database IV, compared to the traditional PC

implementation which achieved 73.5%. In addition,

the processing speed of this accelerator is 8 times that

of PC, with a processing time of 0.01 seconds, while

the processing time of PC is 0.08 seconds. These

results indicate that FPGA based CNN accelerators

have significant advantages in improving

classification accuracy and processing speed.

Dan Liu et al. proposed a discrete wavelet

transform (DWT) system based on FPGA for fea-ture

extraction of EEG signals (Liu et al., 2019). The

system achieves sub-band decomposition of signals

through multi-level DWT. Experimental results show

that the system outperforms traditional software

implementations in terms of accuracy and efficiency

in feature extraction. Specifically, when processing

256 data points, the accuracy of feature extraction in

this system reaches 95.2%, while traditional software

implementation achieves 90.5%. In addition, the

processing time of this system is 0.005 seconds, while

traditional software implementation takes 0.02

seconds. These results indicate that FPGA based

DWT systems have significant advantages in

improving the accuracy and efficiency of feature

extraction.

2.2 Application of FPGA in EEG

Signal Classification

Place Classification, which aims to categorize

extracted features, is the final step in EEG signal

processing for user intent recognition. Chih-Wei Feng

et al. proposed an FPGA based Support Vector

Machine (SVM) classifier for classifying EEG

signals (Feng et al., 2014). This classifier achieves

efficient classification by optimizing the parameters

of SVM. Experimental results show that this classifier

outperforms traditional software implementations in

terms of classification accuracy and processing speed.

Specifically, the classifier achieved an average

classification accuracy of 85.3% when processing

dataset I in BCI Competition database IV, compared

to 80.5% achieved by traditional software

implementations. In addition, the processing time of

this classifier is 0.003 seconds, while traditional

software implementation takes 0.03 seconds. These

results indicate that SVM classifiers based on FPGA

have significant advantages in im-proving

classification accuracy and processing speed.

Figure 2. (a) Rawsignals. (b) Reconstructed signals (Liu et al., 2019).

Apart from classification algorithms, efficient

signal transmission is also crucial. Dan Liu et al.

proposed a multi-channel EEG signal compression

sensing system based on FPGA for EEG signal

transmission in wireless body area networks

(WBAN) (Liu et al., 2019). The system achieves

Research on Real Time EEG Signal Processing Chip Based on FPGA

efficient signal compression and reconstruction

through binary permutation block diagonal matrix

(BPBD). As shown in Figure 2, experimental results

show that the system outperforms traditional

compression methods in terms of compression ratio

and reconstruction accuracy. Specifically, when the

compression ratio of the system is 2, the signal-to-

noise ratio (SNR) of sig-nal reconstruction is

21.74dB, while the SNR of traditional compression

methods is 18.5dB. In addition, the processing time

of this system is 0.002 seconds, while the processing

time of traditional compression methods is 0.01

seconds. These results indicate that FPGA based

compression sensing systems have significant

advantages in improving compression ratio and

reconstruction accuracy.

2.3 Comprehensive Application of

FPGA in EEG Signal Processing

Besides individual applications, FPGA has also made

significant progress in comprehensive EEG signal

processing. Hendrik Wöhrle et al. proposed a real-

time compressed sensing and classification system

based on FPGA for EEG signal processing in wireless

body area networks (Wöhrle et al., 2017). The system

integrates preprocessing, feature extraction, and

classification modules, and achieves efficient signal

processing by optimizing the parameters of each

module. The exper-imental results show that the

system achieves an average classification accuracy of

88.6% with a processing time of 0.005 seconds when

processing multi-channel EEG signals. In comparison,

traditional PC implementation only reaches 83.5%

accuracy and requires 0.05 seconds for processing.

These results indicate that FPGA based integrated

systems have significant ad-vantages in improving

classification accuracy and processing speed.

Figure 3. Pearson correlation analysis in the alpha band.

Pearson correlation analysis in the beta band (Bian et al.,

2018).

Yan Bian et al. proposed a multimodal EEG signal

processing system based on FPGA for real-time

monitoring and classification of multiple EEG signals

(Bian et al., 2018). The system integrates EEG, EMG,

and P300 signal processing modules, and achieves

efficient signal processing by optimizing the

parameters of each module. As shown in Figure 3, the

experimental results show that the system has an

average classification accuracy of 90.2% and a

processing time of 0.008 seconds when processing

multimodal signals, while the traditional PC

implementation has a classification accuracy of

85.5% and a processing time of 0.08 seconds. These

results indicate that FPGA based multimodal systems

have significant advantages in improving

classification accuracy and processing speed.

Despite the significant progress of FPGA in EEG

signal processing, several specific technical

challenges remain to be addressed. Firstly, the

resource utilization and power consumption of FPGA

are still key indicators that need to be optimized.

Future research will focus on develop-ing more

efficient algorithms and architectures to further

improve resource utilization and re-duce power

consumption. Secondly, the application of FPGA in

multimodal signal processing still needs further

exploration. Future research will focus on developing

advanced multimodal signal processing systems to

enhance classification accuracy and processing

efficiency. Final-ly, the application of FPGA in real-

time EEG signal processing still needs further

expansion. Future research will focus on developing

more portable and practical FPGA systems to meet

the needs of different application scenarios.

3 CONCLUSION

This review provides a comprehensive overview of

the state-of-the-art in real-time EEG signal

processing using FPGAs. EEG signals, characterized

by low amplitude and high noise levels, require

sophisticated processing techniques for meaningful

information extraction. Traditional software-based

methods, with their sequential processing nature and

high power demands, are increasingly inadequate for

real-time application. FPGAs, with their parallel

architecture and reconfigurability, offer a

transformative solution. This review highlights key

applications of FPGAs in EEG signal processing,

including artifact removal in preprocessing, wave-let-

BEFS 2025 - International Conference on Biomedical Engineering and Food Science

based feature extraction, SVM-based classification,

and integrated system design. Comparative analysis

shows significant improvements in processing speed,

resource utilization, and classification accuracy over

traditional methods. However, challenges remain in

optimizing re-source utilization, expanding

multimodal applications, and enhancing portability.

The future of FPGA-based EEG signal processing

holds great promise for advancing both clinical and

research applications. As technology continues to

evolve, the development of more efficient algorithms

and architectures will be crucial. These advancements

will further optimize resource utilization and reduce

power consumption, making FPGA-based systems

even more practical for portable and wearable

devices. Additionally, the integration of multi-modal

signal processing will enhance classification

accuracy. Combining EEG with other physiological

signals, such as EMG and ECG, provides a more

comprehensive understanding of brain and body

interactions. Moreover, the expansion of real-time

processing capabilities will enable seamless

integration into wireless body area networks

(WBANs), facilitating continuous monitoring and

analysis. Future research should also focus on

improving the accessibility and user-friendliness of

FPGA-based systems, ensuring they can be easily

adopted by re-searchers and clinicians without

extensive hardware expertise. Overall, the continuous

innovation in FPGA technology will drive the

development of more efficient, versatile, and

practical solutions for EEG signal processing, from

portable medical devices to advanced research tools,

ultimately enhancing our ability to decode brain

activity and improve healthcare outcomes.

REFERENCES

Aravind, M., & Suresh Babu, S. 2016. Embedded

Implementation of Brain Computer Interface using

FPGA. 2016 International Conference on Emerging

Technological Trends [ICETT]:1-6.

Bian, Yan, Qi, Hongzhi, Zhao, Li, Ming, Dong, Guo, Tong,

& Fu, Xing. 2018. Improvements in event-related

desynchronization and classification performance of

motor imagery using instructive dynamic guidance and

complex tasks. Computers in Biology and Medicine:1-

12.

Feng, Chih-Wei, Hu, Ting-Kuei, Chang, Jui-Chung, &

Fang, Wai-Chi. 2014. A Reliable Brain Computer

Interface Implemented on an FPGA for a Mobile

Dialing System. IEEE Transactions on Biomedical

Circuits and Systems:654-657.

Lin, Jzau-Sheng, & Huang, Sun-Ming. 2013. An FPGA-

Based Brain-Computer Interface for Wireless Electric

Wheelchairs. Applied Mechanics and Materials Vols.

284-287:1616-1621.

Liu, Dan, Wang, Qisong, Zhang, Yan, Liu, Xin, Lu,

Jingyang, & Sun, Jinwei. 2019. FPGA-based real-time

compressed sensing of multichannel EEG signals for

wireless body area networks. Biomedical Signal

Processing and Control 49:221-230.

Ma, Xunguang, Zheng, Wenkai, Peng, Zujian, & Yang,

Jimin. 2019. FPGA-Based Rapid

Electroencephalography Signal Classification System.

2019 IEEE 11th International Conference on Advanced

Infocomm Technology:223-227.

Shyu, Kuo-Kai, et al. 2013. Adaptive SSVEP-Based BCI

System With Frequency and Pulse Duty-Cycle Stimuli

Tuning Design. IEEE Transactions on Neural Systems

and Rehabilitation Engineering 21(5):697-703.

Sundaram, Kalyana, Marichamy, & Pradeepa. 2016. FPGA

Based Filters for EEG Pre-processing. 2016 Second

International Conference on Science Technology

Engineering And Management (ICONSTEM):572-577.

Tamilarasi, S., & Sundararajan, J. 2018. FPGA based

seizure detection and control for brain computer

interface. Cluster Computing:1-20.

Wahalla, Marc-Nils, Payá Vayá, Guillermo, & Blume,

Holger. 2020. CereBridge: An Efficient, FPGA-based

Real-Time Processing Platform for True Mobile Brain-

Computer Interfaces. IEEE Transactions on Biomedical

Engineering:4046-4050.

Wöhrle, Hendrik, Tabie, Marc, Kim, Su Kyoung, Kirchner,

Frank, & Kirchner, Elsa Andrea. 2017. A Hybrid

FPGA-Based System for EEG-and EMG-Based Online

Movement Prediction. Sensors 17(7):1552.

Zhao, Dongsheng, Jiang, Jiang, Wang, Chang, Lu,

Baoliang, & Zhu, Yongxin. 2015. FPGA

Implementation of FastICA Algorithm for On-line EEG

Signal Separation. Communications in Computer and

Information Science 491:59-68.

Research on Real Time EEG Signal Processing Chip Based on FPGA