Decoding Of Hand Gestures Information Based On Corticomuscular Coherence

ObjectiveRecovery of upper limb disorder is directly related to the quality of stroke patients. It is the functional connectivity between motor cortex and muscle that directly related to the rehabilitation of the dysfunction in upper limb. It’s well known that upper limb movement relies on the brain’s control, and there is interaction between the cortex and the muscles during exercise, while neuromuscular activity status can be detected by EEG-EMG coherence analysis. In order to achieve upper limb motion information decoding, and provide experimental support for exploring the upper limb motor function reconstruction mechanism. This article adopts the method of EEG-EMG collaborative analysis, and records EEG from the motor cortex and EMG from forearm surface at the same time during four typical state of hand motions. Coherence analysis can be applied to research the characteristics between cortex and muscle under different movement patterns.Methods1. Experimental paradigm and data acquisition, preprocessingThe experimental data was recorded by multiple physiological signal recording system, we get the signals which consist of9channels EEG of F3F4Fz C3C4Cz P3P4Pz from motor cortex and4channels EMG of FD ED FCU ECR from right forearm, during4groups of hand motions including flexor digitorum, extensor digitorum.wrist flexion, and wrist extension in10healthy subjects according to the paradigm. Data of each subject in each group action was collected three times, a total of30groups. Signals were separated into two parts by tag of flexion and extension movements based on the paradigm. And signals were split into movement preparation signals and movement execution signals based on the tag of preparation (tag of high level sustained0.5s from the tag of flexion or extension) and execution (tag of low level sustained2.5s from the tag of preparation).The9-channel EEG signals were divided into five bands:8-band (0.5-4Hz),9-band (4-8Hz), a-band (8-13Hz),(β-band (13-30Hz), y-band (30-60Hz). The data were processed through baseline drift correction and power frequency filtering.2. EEG, sEMG coherence analysis and pattern recognitionThe method of coherence analysis and statistical analysis were used to analyze EEG of each band and EMG of4muscles (FD, ED, FCU, ECR) which were recorded from4groups of hand motions from two aspects of the motion (movement preparation phase and movement execution phase). The motions were recognized using support vector machine based on the feture extracted from the coherence coefficients.Results1. Results of coherence analysis1) Movement preparation phaseIn the β-band, the coherence coefficients between EEG of C3P3Pz and FD is greater than that between EEG of C3P3Pz and ED in the right hand flexor digitorum movement(P<0.05). The coherence coefficients between EEG of C3P3Pz and ED is greater than that between EEG of C3P3Pz and FD in the right hand extensor digitorum movement(P<0.05). The coherence coefficients between EEG(all channels) and FCU is greater than that between EEG and ECR in the right hand wrist flexion movement(P<0.05). The coherence coefficients between EEG of F3C3Cz C4P4and ECR is greater than that between EEG of F3C3Cz C4P4and FCU in the right hand wrist extension movement(P<0.05).In the γ-band, the coherence coefficients between EEG of Fz C3and FD is greater than that between EEG of Fz C3and ED in the right hand flexor digitorum movement(P<0.05). The coherence coefficients between EEG of Fz C3and ED is greater than that between EEG of Fz C3and FD in the right hand extensor digitorum movement(P<0.05). The coherence coefficients between EEG(F3Fz C3Cz C4P3) and FCU is greater than that between EEG and ECR in the right hand wrist flexion movement(P<0.05). The coherence coefficients between EEG of C3Cz P3Pz P4and ECR is greater than that between EEG of C3Cz P3Pz P4and FCU in the right hand wrist extension movement(P<0.05).During the preparation of the movement, C4P4on the right motor cortex exhibit coherence with the right front wrist muscles in the wrist movements.2) Movement execution phaseIn the β-band. the coherence coefficients between C3and FD is greater than that between C3and ED in the right hand flexor digitorum movement(P<0.05). The coherence coefficients between C3and ED is greater than that between C3and FD in the right hand extensor digitorum movement(P<0.05). The coherence coefficients between C3and FCU is greater than that between C3and ECR in the right hand wrist flexion movement(P<0.01). The coherence coefficients between C3and ECR is greater than that between C3and FCU in the right hand wrist extension movement(P<0.01). 2. Results of pattern recognitionThe coherence coefficients between EEG (β-band, γ-band) and EMG were used as characteristic parameter to conduct SVM pattern recognition. The correct rate of flexor digitorum and extensor digitorum is95%, while that of wrist flexion and wrist extension is90%.ConclusionEEG of all channels in β-band and γ-band exhibit coherence with EMG during the preparation of the movement. This indicates that the motor cortex are more active than usual, not only the left brain which controls the right body, but also C4P4on the right brain exhibit coherence with the right front wrist muscles in the wrist movements.During the execution of the movement, the functional connectivity of upper limb muscles is directly come from the contralateral motor cortex, while the motor cortex region on the muscle side has no significant coherence.EEG in β-band and γ-band shows significant coherence with sEMG of upper limb muscles during hand movements. The correct rate of both hand grip and wrist movements is over90%. It can serve as the basis for the information decoding of hand movement.