Conselho Nacional de Desenvolvimento Científico e Tecnológico / The detection and classification of EMG motor unit action potentials (MUAP) is an important tool in the study of the neuromuscular system, allowing for a number os applications, such as the diagnoses of motor disorders. However, although there are several methods described in the literature to perform such actions, the majority relies on complex algorithms and specific instrumentation. Depending on the system, the computational cost or the detection mechanism, sometimes involving electrode arrays, may limit its use in clinical applications, biofeedback or embedded systems for controlling artificial prostheses. Another important issue is the detection and classification of firing MUAPs in signals with low signal to noise ratio (SNR). A method capable of operating with low SNR is paramount for applications, such as the use of electromyography in human machine interfaces (HMI), where the positioning and fixation of the electrodes may be performed by a non-trained user, and the signal can be contaminated by high levels of electromagnetic interference. As a solution for such problems, two complementary methods were proposed: the first (MD-FC) is based on the use of banks of matched filters for detection and classification MUAPs in EMG signals, whereas the second (MAD-FC) is proposed as an improvement from the first, aiming situations with a high probability of overlapping firing MUAPs. The proposed methods sought to achieve those goals without an excessive increase in computational cost, treating signals with variable noise levels and considering the overlapping of MUAPs. The results showed that the MD-FC system is able to accurately detect 96% of isolated MUAPs in signals with SNR of 10 dB and up to 10 active motor units. However, the performance is reduced in the presence of high levels of overlapping MUAPs, as expected. The second method (MAD-FC) was designed to improve the detection of overlapping MUAPs. The results showed that the MAD-FC is able to detect and classify firing MUAPs in signals with up 10 active motor units and SNR of 20 dB at rates of success higher than 79.80%, on average. When the SNR is decreased to 10dB the rates of success reach at least 75.19%, on average (even in this case with a high percentage of overlapping). In general, the MAD-FC showed rates of success around 20% better than the MD-FC method. Both methods are quite efficient when using computational resources. They were created in order to process EMG windows of 50 milliseconds in less than 5 milliseconds, when using a standard desktop computer. This feature allows their use in applications requiring MUAPs detection and classification in real time. / A detecção e classificação dos potenciais de ação de unidade motora (PAUMs) de sinais EMG é uma ferramenta importante no estudo do sistema neuromuscular. A partir de informações dessa classificação é possível diagnosticar distúrbios motores. Entretanto, apesar de existirem diversas propostas na literatura para executar tais ações, a grande maioria depende de algoritmos complexos e instrumentação específica. Dependendo do sistema, o custo computacional ou o mecanismo de captação envolvendo, matrizes de eletrodos, pode limitar sua utilização em aplicações clínicas, biofeedback ou em sistemas embarcados para controle de próteses. Outra questão importante consiste na detecção e classificação de disparos em sinais com baixa relação sinal ruído (SNR). Um método capaz de operar em sinais com baixa SNR é interessante em aplicações onde não se pode controlar completamente a coleta do sinal. Como exemplo, podemos citar aplicações da eletromiografia em interfaces homem máquina (IHM), onde o posicionamento dos eletrodos pode ser realizado por um usuário com pouco treinamento e o ambiente pode conter alto nível de interferência eletromagnética, diminuindo a SNR do sinal captado. Como solução para tais problemas, foram propostas duas metodologias complementares: a primeira delas (MD-FC) se baseia no uso de bancos de filtros casados para detecção e classificação de PAUMs em sinais EMG, enquanto a segunda (MAD-FC) é uma proposta de aprimoramento da primeira para situações com altas probabilidades de sobreposição de disparos de MUAPs. As metodologias propostas buscaram atingir aqueles objetivos sem um aumento excessivo no custo computacional, tratando sinais com níveis variados de ruídos e considerando a questão de sobreposição de PAUMs, comuns em sinais EMG. Os resultados demonstraram que o sistema MD-FC é capaz de detectar disparos isolados com precisão de 96% em média para relação sinal ruído de 10 dB com até 10 unidades motoras ativas, porém seu é desempenho diminuído na presença de altos níveis de sobreposição de PAUMS. O segundo MAD-FC que foi elaborado de forma a aprimorar a detecção sobre ondas sobrepostas, e é capaz de detectar e classificar os disparos de sinais com até 10 unidades motoras ativas com taxa de classificação correta maior do que 79,80% em média e com SNR de 20 dB. Para sinais com SNR de 10 dB esse valor é de 75,19% em média. Em geral, o método MAD-FC apresentou performance superior ao MD-FC em cerca de 20%. Os dois métodos são bastante eficientes no uso de recursos computacionais. Eles foram criadas de forma a analisar os sinais EMG em janelas de 50 milissegundos em menos de 5 milissegundos a partir de um computador desktop padrão, o que permite sua utilização em sistemas de detecção e classificação de PAUMs em tempo real. / Doutor em Ciências

Electromyographic (EMG) signal decomposition is the process of resolving an EMG signal into its component motor unit potential trains (MUPTs). The extracted MUPTs can aid in the diagnosis of neuromuscular disorders and the study of the neural control of movement, but only if they are valid trains. Before using decomposition results and the motor unit potential (MUP) shape and motor unit (MU) firing pattern information related to each active MU for either clinical or research purposes the fact that the extracted MUPTs are valid needs to be confirmed. The existing MUPT validation methods are either time consuming or related to operator experience and skill. More importantly, they cannot be executed during automatic decomposition of EMG signals to assist with improving decomposition results. To overcome these issues, in this thesis the possibility of developing automatic MUPT validation algorithms has been explored. Several methods based on a combination of feature extraction techniques, cluster validation methods, supervised classification algorithms, and multiple classifier fusion techniques were developed. The developed methods, in general, use either the MU firing pattern or MUP-shape consistency of a MUPT, or both, to estimate its overall validity. The performance of the developed systems was evaluated using a variety of MUPTs obtained from the decomposition of several simulated and real intramuscular EMG signals. Based on the results achieved, the methods that use only shape or only firing pattern information had higher generalization error than the systems that use both types of information. For the classifiers that use MU firing pattern information of a MUPT to determine its validity, the accuracy for invalid trains decreases as the number of missed-classification errors in trains increases. Likewise, for the methods that use MUP-shape information of a MUPT to determine its validity, the classification accuracy for invalid trains decreases as the within-train similarity of the invalid trains increase. Of the systems that use both shape and firing pattern information, those that separately estimate MU firing pattern validity and MUP-shape validity and then estimate the overall validity of a train by fusing these two indices using trainable fusion methods performed better than the single classifier scheme that estimates MUPT validity using a single classifier, especially for the real data used. Overall, the multi-classifier constructed using trainable logistic regression to aggregate base classifier outputs had the best performance with overall accuracy of 99.4% and 98.8% for simulated and real data, respectively. The possibility of formulating an algorithm for automated editing MUPTs contaminated with a high number of false-classification errors (FCEs) during decomposition was also investigated. Ultimately, a robust method was developed for this purpose. Using a supervised classifier and MU firing pattern information provided by each MUPT, the developed algorithm first determines whether a given train is contaminated by a high number of FCEs and needs to be edited. For contaminated MUPTs, the method uses both MU firing pattern and MUP shape information to detect MUPs that were erroneously assigned to the train. Evaluation based on simulated and real MU firing patterns, shows that contaminated MUPTs could be detected with 84% and 81% accuracy for simulated and real data, respectively. For a given contaminated MUPT, the algorithm on average correctly classified around 92.1% of the MUPs of the MUPT. The effectiveness of using the developed MUPT validation systems and the MUPT editing methods during EMG signal decomposition was investigated by integrating these algorithms into a certainty-based EMG signal decomposition algorithm. Overall, the decomposition accuracy for 32 simulated and 30 real EMG signals was improved by 7.5% (from 86.7% to 94.2%) and 3.4% (from 95.7% to 99.1%), respectively. A significant improvement was also achieved in correctly estimating the number of MUPTs represented in a set of detected MUPs. The simulated and real EMG signals used were comprised of 3–11 and 3–15 MUPTs, respectively.

Electromyographic (EMG) signal decomposition is the process of resolving an EMG signal into its component motor unit potential trains (MUPTs). The extracted MUPTs can aid in the diagnosis of neuromuscular disorders and the study of the neural control of movement, but only if they are valid trains. Before using decomposition results and the motor unit potential (MUP) shape and motor unit (MU) firing pattern information related to each active MU for either clinical or research purposes the fact that the extracted MUPTs are valid needs to be confirmed. The existing MUPT validation methods are either time consuming or related to operator experience and skill. More importantly, they cannot be executed during automatic decomposition of EMG signals to assist with improving decomposition results. To overcome these issues, in this thesis the possibility of developing automatic MUPT validation algorithms has been explored. Several methods based on a combination of feature extraction techniques, cluster validation methods, supervised classification algorithms, and multiple classifier fusion techniques were developed. The developed methods, in general, use either the MU firing pattern or MUP-shape consistency of a MUPT, or both, to estimate its overall validity. The performance of the developed systems was evaluated using a variety of MUPTs obtained from the decomposition of several simulated and real intramuscular EMG signals. Based on the results achieved, the methods that use only shape or only firing pattern information had higher generalization error than the systems that use both types of information. For the classifiers that use MU firing pattern information of a MUPT to determine its validity, the accuracy for invalid trains decreases as the number of missed-classification errors in trains increases. Likewise, for the methods that use MUP-shape information of a MUPT to determine its validity, the classification accuracy for invalid trains decreases as the within-train similarity of the invalid trains increase. Of the systems that use both shape and firing pattern information, those that separately estimate MU firing pattern validity and MUP-shape validity and then estimate the overall validity of a train by fusing these two indices using trainable fusion methods performed better than the single classifier scheme that estimates MUPT validity using a single classifier, especially for the real data used. Overall, the multi-classifier constructed using trainable logistic regression to aggregate base classifier outputs had the best performance with overall accuracy of 99.4% and 98.8% for simulated and real data, respectively. The possibility of formulating an algorithm for automated editing MUPTs contaminated with a high number of false-classification errors (FCEs) during decomposition was also investigated. Ultimately, a robust method was developed for this purpose. Using a supervised classifier and MU firing pattern information provided by each MUPT, the developed algorithm first determines whether a given train is contaminated by a high number of FCEs and needs to be edited. For contaminated MUPTs, the method uses both MU firing pattern and MUP shape information to detect MUPs that were erroneously assigned to the train. Evaluation based on simulated and real MU firing patterns, shows that contaminated MUPTs could be detected with 84% and 81% accuracy for simulated and real data, respectively. For a given contaminated MUPT, the algorithm on average correctly classified around 92.1% of the MUPs of the MUPT. The effectiveness of using the developed MUPT validation systems and the MUPT editing methods during EMG signal decomposition was investigated by integrating these algorithms into a certainty-based EMG signal decomposition algorithm. Overall, the decomposition accuracy for 32 simulated and 30 real EMG signals was improved by 7.5% (from 86.7% to 94.2%) and 3.4% (from 95.7% to 99.1%), respectively. A significant improvement was also achieved in correctly estimating the number of MUPTs represented in a set of detected MUPs. The simulated and real EMG signals used were comprised of 3–11 and 3–15 MUPTs, respectively.