Method for improving the signal to noise ratio of a wave form
Inventors
Siegel, David • Ly, Canh • Lau, Troy • Hairston, William D.
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-10524733-B2
Publication Date
2020-01-07
Expiration Date
2034-04-21
Abstract
A method for improving the signal to noise ratio of an EEG signal in which a wavelet packet decomposition having a plurality of levels is first applied to a time slice of the EEG signal. A default signal is set to the first wavelet packet and a default peak response is then calculated for the first wavelet node. An update signal is set to the default signal combined with another of the wavelet nodes and an update peak response signal is then calculated of the update signal. If the update peak response signal exceeds the default peak response, the default peak response is set equal to the update peak response and the default signal is set equal to the update signal. Otherwise, the value of the current node is set to zero which effectively eliminates the signal data of the current wavelet node. These steps are reiterated for all of the wavelet nodes and, thereafter, a composite waveform of the EEG signal is reconstructed from the non-zero wavelet nodes.
Core Innovation
The invention is a processor implemented method for improving the signal to noise ratio of single trial electroencephalogram (EEG) signals, specifically adapting to different sequential time segments or time slices over the entire duration of the EEG signal. The method begins by acquiring a single trial of the EEG response such as the P300 signal, then applying a wavelet packet decomposition of N levels, typically seven, to form wavelet nodes corresponding to adjacent frequency pass bands. Each wavelet node contains coefficients indicating the magnitude and direction of the EEG signal in that frequency range.
The method sets a default signal to the first wavelet node and calculates a default peak response based on a ratio of the peak response to the root mean square of the default signal. It then combines the default signal with each successive wavelet node, recalculates an update peak response, and compares this value to the default peak response. If the update peak response is greater, the default signal and peak response are updated; if not, the current wavelet node's coefficient is set to zero, effectively removing its influence. This iterative process is repeated for all wavelet nodes and repeated for multiple sequential time slices, allowing the method to adaptively improve signal to noise ratio across time-varying EEG noise.
Claims Coverage
The patent claims include two independent claims describing methods for improving signal to noise ratio in single trial EEG waveforms through wavelet decomposition and adaptive filtering over time.
Adaptive wavelet packet decomposition for EEG signals
Applying a wavelet packet decomposition to the single trial EEG waveform to form adjacent wavelet nodes, each corresponding to a frequency pass band with coefficients representing the magnitude of the P300 response.
Constructing and updating signals based on peak response ratios
Setting a default signal from the first wavelet node and calculating a default peak response defined as the ratio of peak magnitude to root-mean-square (RMS), then iteratively updating the signal by combining with subsequent wavelet nodes and recalculating the update peak response.
Selective elimination of wavelet nodes to improve signal to noise ratio
If the update peak response does not exceed the default, the next wavelet node’s coefficients are set to zero, effectively removing noise contributions, while if it does, the default signal and peak response are updated to the new values.
Reconstructing an improved EEG waveform from updated wavelet nodes
After iteration through all wavelet nodes, reconstructing a composite waveform from wavelet nodes with non-zero coefficients to form an improved EEG signal with enhanced signal to noise ratio.
Repeating the method over sequential time slices
Applying the above process iteratively to multiple sequential time segments of the EEG signal to adaptively reduce noise over time.
The inventive features focus on adaptively applying wavelet packet decomposition combined with iterative evaluation of peak response ratios to selectively retain or eliminate frequency components, reconstructing an improved EEG waveform across time slices that enhances signal to noise ratio for single trial EEG signals.
Stated Advantages
Automatically adapts noise filtering to different sequential time segments of the EEG signal.
Significantly improves the signal to noise ratio despite variations of signal and noise across the analysis period.
Eliminates wavelet nodes that do not contribute to improving the EEG signal, effectively reducing noise.
Improves identification and analysis of important EEG responses such as the P300 signal.
Documented Applications
Monitoring EEG signals in hospitals and medical facilities.
Monitoring EEG signals of warriors in military combat arenas.
Analyzing P300 responses in EEG signals for reaction assessments to stimuli.
Interested in licensing this patent?