EMG and ENG-envelope pattern recognition for prosthetic hand control

• Published in: Journal of neuroscience methods Vol. 311; pp. 38 - 46
• Main Authors: Noce, Emiliano, Dellacasa Bellingegni, Alberto, Ciancio, Anna Lisa, Sacchetti, Rinaldo, Davalli, Angelo, Guglielmelli, Eugenio, Zollo, Loredana
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.01.2019
• Subjects: Algorithms; Artificial Limbs; Electromyography - methods; EMG signals; Hand - innervation; Hand - physiopathology; Humans; Male; Median Nerve - physiopathology; Movement - physiology; Muscle, Skeletal - innervation; Muscle, Skeletal - physiopathology; Neural signals; Neuroprosthetics; Pattern recognition; Pattern Recognition, Automated - methods; Signal Processing, Computer-Assisted; Spike sorting; Support Vector Machine; Ulnar Nerve - physiopathology; Neuroprosthetics; Spike sorting; Pattern recognition; EMG signals; Neural signals
• ISSN: 0165-0270; 1872-678X; 1872-678X
• DOI: 10.1016/j.jneumeth.2018.10.004

•A novel mathematical approach is proposed to process neural (ENG) signals.•ENG envelope is used to classify real neural data recorded from a human amputee.•A classification algorithm based on ENG envelope outperforms the spike sorting ones.•ENG envelope is computational efficient and can be used in on-line applications. This paper proposes a new approach for neural control of hand prostheses, grounded on pattern recognition applied to the envelope of neural signals (eENG). The ENG envelope was computed by taking into account the amplitude and the occurrence of the spike in the neural recording. A pattern recognition algorithm applied on muscular signals was defined as a reference and a comparative analysis with traditionally adopted Spike Sorting Algorithms (SSA) for neural signals has been carried out. Method validation was divided in two parts: firstly, neural signals recorded from one amputee subject through intraneural electrodes were offline analyzed to discriminate between the two performed gestures; secondly, algorithm performance decay with the increase of the number of classes was studied through synthetic data. An accuracy of 98.26% with real data was reached with the pattern recognition applied to eENG. SSA reached an accuracy of 70%. Increasing the number of classes worsens the accuracy of this algorithm. Additionally, computational time for the pattern recognition applied to eENG is very low (32.6 μs for each sample in the data window analyzed). The eENG was proved to be more reliable in decoding the user intention than the SSA algorithm and it is computationally efficient. It was demonstrated that it is possible to apply the well-known techniques of EMG pattern recognition to a conveniently processed neural signal and can pave the way to the application of neural gesture decoding in upper limb prosthetics.