This paper presents a literature review on pattern recognition of electromyography (EMG) signals and its applications. The EMG technology is introduced and the most relevant aspects for the design of an EMG-based system are highlighted, including signal acquisition and filtering. EMG-based systems have been used with relative success to control upper-and lower-limb prostheses, electronic devices and machines, and for monitoring human behavior. Nevertheless, the existing systems are still inadequate and are often abandoned by their users, prompting for further research. Besides controlling prostheses, EMG technology is also beneficial for the development of machine learning-based devices that can capture the intention of able-bodied users by detecting their gestures, opening the way for new human-machine interaction (HMI) modalities. This paper also reviews the current feature extraction techniques, including signal processing and data dimensionality reduction. Novel classification methods and approaches for detecting non-trained gestures are discussed. Finally, current applications are reviewed, through the comparison of different EMG systems and discussion of their advantages and drawbacks.INDEX TERMS EMG, human-machine interaction, pattern classification, regression.

International audienceRecent algebraic parametric estimation techniques (see \cite{garnier,mfhsr}) led to point-wise derivative estimates by using only the iterated integral of a noisy observation signal (see \cite{num0,num}). In this paper, we extend such differentiation methods by providing a larger choice of parameters in these integrals: they can be reals. For this, %as in \cite{num0,num}, the extension is done via a truncated Jacobi orthogonal series expansion. Then, the noise error contribution of these derivative estimations is investigated: after proving the existence of such integral with a stochastic process noise, their statistical properties (mean value, variance and covariance) are analyzed. In particular, the following important results are obtained: \begin{description} \item[$a)$] the bias error term, due to the truncation, can be reduced by tuning the parameters, \item[$b)$] such estimators can cope with a large class of noises for which the mean and covariance are polynomials in time (with degree smaller than the order of derivative to be estimated), \item[$c)$] the variance of the noise error is shown to be smaller in the case of negative real parameters than it was in \cite{num0,num} for integer values. \end{description} Consequently, these derivative estimations can be improved by tuning the parameters according to the here obtained knowledge of the parameters' influence on the error bounds