Related Experiment Video
Updated: Jan 14, 2026
08:37
Focused Ultrasound Induced Blood-Brain Barrier Opening for Targeting Brain Structures and Evaluating Chemogenetic Neuromodulation
Published on: December 22, 2020
4.2K
Neuroanatomy-Informed Brain-Machine Hybrid Intelligence for Robust Acoustic Target Detection
Jianting Shi1, Jiaqi Wang1, Weijie Fei1
1School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, China.
Cyborg and Bionic Systems (Washington, D.C.)
|October 20, 2025
Summary
This study introduces a novel brain-computer interface (BCI) for sound target detection (STD), improving robustness in noisy conditions. A hybrid neuro-acoustic system enhances detection accuracy and generalizes to new sound types.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Acoustic Sensing
Background:
- Automated sound target detection (STD) methods lack robustness and generalization, especially in low signal-to-noise ratio (SNR) environments or with novel sound categories.
- Existing systems struggle with reliability and accuracy in complex acoustic scenes, limiting their real-world applicability.
Purpose of the Study:
- To develop a robust and generalizable sound target detection (STD) method by integrating brain-computer interface (BCI) technology with conventional acoustic sensing.
- To enhance the accuracy and interpretability of STD in complex auditory environments using neural responses.
- To overcome the limitations of standalone BCI systems, such as high false alarm rates, through a hybrid fusion strategy.
Main Methods:
- Proposed a Triple-Region Spatiotemporal Dynamics Attention Network (Tri-SDANet), an electroencephalogram (EEG) decoding model incorporating neuroanatomical priors from EEG source analysis.
- Developed an adaptive confidence-based brain-machine fusion strategy to combine BCI and acoustic detection model decisions.
- Conducted experiments with 16 participants to validate the neuro-acoustic fusion approach.
Main Results:
- The Tri-SDANet achieved state-of-the-art performance in neural decoding under complex acoustic conditions.
- The hybrid system demonstrated reliable detection performance at low SNR levels and remarkable generalization to unseen target classes.
- Source-level EEG analysis revealed distinct brain activation patterns linked to target perception, validating the model's design.
Conclusions:
- Pioneered a neuro-acoustic fusion paradigm for robust and generalizable sound target detection (STD).
- The integrated system effectively merges neural perception and acoustic feature learning, offering a significant advancement over existing methods.
- This approach provides a promising, generalizable solution for real-world acoustic sensing applications by leveraging noninvasive neural signals and artificial intelligence.
