Abstract

Understanding the decision-making processes behind voluntary and involuntary motor actions remains a central challenge in neuroscience. Conventional EEG markers such as event-related desynchronization/synchronization (ERD/ERS) have provided valuable insights into motor intention but suffer from persistent limitations—most notably the reliance on pre-stimulus baselines. These baselines are highly sensitive to inter-trial variability, transient state changes, and non-stationary noise, making absolute amplitude comparisons unreliable across trials, subjects, and recording sessions.

This thesis introduces cumulative power spectral density (CPSD) in the beta and gamma bands as a robust, adaptive alternative. Unlike ERD/ERS, which requires frequent baseline recalculations (e.g., every 2 seconds from the preceding 0.5 s), CPSD avoids fixed baseline subtraction by using a sliding-window accumulation approach that updates its max–min reference only when a new extreme PSD value is detected. This method preserves temporal dynamics while minimizing the instability introduced by fluctuating baselines.

The approach leverages data-driven, sliding-window extraction of cumulative power spectral density (CPSD) features in the beta and gamma bands, enabling fine-grained characterization of motor intention and cognitive control. Comprehensive analysis across 109 healthy subjects shows that optimal classification performance is achieved with short accumulation windows (0.05–0.20s), particularly around movement cue onset. Beta and gamma CPSD features exhibit distinct temporal and spatial dynamics, with beta generally favoring slightly longer integration windows. These results highlight that motor intention signatures emerge within narrow, task-specific temporal windows, and that beta–gamma CPSD provides a stable and interpretable alternative to traditional ERD/ERS measures for decoding voluntary motor control.

Beyond classification, CPSD revealed coordinated beta–gamma dynamics underlying selective attention, response inhibition, and intention-to-execution transitions, offering a richer, more interpretable representation of motor control processes than conventional ERD/ERS measures.

The results establish that the cumulative beta--gamma power within task-specific windows serves as a core neural marker for voluntary motor control. This method allows for precise, automated classification of ME/MI states by referencing adaptive CPSD thresholds, providing a simultaneous measure of motor intention strength. Their interplay encodes the core neural processes for voluntary control. These findings position CPSD as a paradigm-shifting neural feature for next-generation real-time brain–computer interfaces (BCIs) and clinically relevant neuro-technologies, providing a principled foundation for decoding conscious versus unconscious motor actions. By replacing fixed cutoffs with adaptive, data-driven thresholds, CPSD advances the precision of motor intention classification and deepens our understanding of brain dynamics in human motor control, potentially redefining EEG-based motor intention research.