Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
97 tokens/sec
GPT-4o
53 tokens/sec
Gemini 2.5 Pro Pro
44 tokens/sec
o3 Pro
5 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Brain-inspired Computing Based on Deep Learning for Human-computer Interaction: A Review (2312.07213v4)

Published 12 Dec 2023 in cs.AI

Abstract: The continuous development of artificial intelligence has a profound impact on biomedicine and other fields, providing new research ideas and technical methods. Brain-inspired computing is an important intersection between multimodal technology and biomedical field. Focusing on the application scenarios of decoding text and speech from brain signals in human-computer interaction, this paper presents a comprehensive review of the brain-inspired computing models based on deep learning (DL), tracking its evolution, application value, challenges and potential research trends. We first reviews its basic concepts and development history, and divides its evolution into two stages: recent machine learning and current deep learning, emphasizing the importance of each stage in the research of brain-inspired computing for human-computer interaction. In addition, the latest progress of deep learning in different tasks of brain-inspired computing for human-computer interaction is reviewed from five perspectives, including datasets and different brain signals, and the application of key technologies in the model is elaborated in detail. Despite significant advances in brain-inspired computational models, challenges remain to fully exploit their capabilities, and we provide insights into possible directions for future academic research. For more detailed information, please visit our GitHub page: https://github.com/ultracoolHub/brain-inspired-computing.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (6)
  1. Bihui Yu (16 papers)
  2. Sibo Zhang (15 papers)
  3. Lili Zhou (1 paper)
  4. Jingxuan Wei (21 papers)
  5. Linzhuang Sun (18 papers)
  6. Liping Bu (3 papers)

Summary

  • The paper provides a comprehensive review of how brain-inspired computing models leverage deep learning to enhance the decoding of neural signals and cognitive tasks.
  • It demonstrates the use of deep learning techniques to analyze EEG signals for emotion classification and text translation, underscoring significant methodological advances.
  • It discusses challenges such as limited datasets and high computational demands, while advocating future research on non-invasive methods and spiking neural networks.

Evolution of Brain-Inspired Computing Models

Introduction to Brain-Inspired Computing

Brain-inspired computing refers to a kind of machine intelligence that draws inspiration from the neural mechanisms and cognitive behavior of the human brain. It employs computational modeling and is realized through both software and hardware cooperation. The human brain, a highly efficient organ, accomplishes complex tasks using only a fraction of the body’s energy. Brain-inspired systems aim to replicate human cognitive behavior and intelligence levels.

Advances in Machine Learning and Deep Learning

Machine learning and deep learning have seen significant application in brain-inspired computing. The paper of this field encompasses ML and deep learning (DL) techniques for analyzing brain signals and developing robust predictive models.

Applications of Deep Learning Models

Recent advancements have demonstrated the power of DL in various tasks related to brain-inspired computing. Researchers made strides in the domain of decoding brain signals for tasks like predicting word sequences and classifying emotions through EEG signals. There have been impressive developments in EEG-related applications and techniques, such as the cross-modal cloze task and EEG to text translation. By combining brain physiological data with DL models, applications are beginning to emerge that can read comprehension directly from brain activity.

Challenges and Future Directions

Despite progress, there are challenges that need to be addressed to make full use of brain-inspired computing's capabilities. These include the need for high-quality, extensive datasets, which are expensive and challenging to collect. As most development depends on invasive recording technology, finding ways to enhance non-invasive methods without sacrificing accuracy is crucial.

Furthermore, DL models require substantial computational resources, limiting their accessibility. Researchers are encouraged to consider more efficient training algorithms and the use of lightweight deployment models to overcome these hurdles. The accuracy and generalization performance of models also need improvement, as diversity among individual brainwave patterns and across datasets can vary significantly.

Looking ahead, future research directions are promising and could change the course of AI and brain simulation. One potential area is exploring applications for non-invasive brain-inspired computing in fields such as medicine and psychology. Also, the integration of Spiking Neural Networks (SNNs), which offer a closer simulation of brain function, could provide a more robust framework for brain-inspired AI, with advantages like enhanced biological plausibility and energy efficiency.

In closing, this paper aims to provide a comprehensive review of the development of brain-inspired computing, offering an insightful reference for future academic research. The GitHub repository linked in the paper can be accessed for more detailed information and resources.