Context. EMG (Electromyographic) signal is a response of a neuromuscular system for an electrical stimulus generated either by brain or by spinal cord. This thesis concerns the subject of onset detection in the context of a muscle activity. Estimation is based on an EMG signal observed during a muscle activity.

Objectives. The aim of this research is to propose new onset estimation algorithms and compare them with solutions currently existing in the academia. Two benchmarks are being considered to evaluate the algorithms’ results- a muscle torque signal synchronized with  an EMG signal and a specialist’s assessment. Bias, absolute value of a mean error and standard deviation are the criteria taken into account.

Methods. The research is based on EMG data collected in the physiological laboratory at Wroclaw University of Physical Education. Empty samples were cut off the dataset. Proposed estimation algorithms were constructed basing on the EMG signal analysis and review on state of the art solutions. In order to collate them with existing solutions a simple comparison have been conducted.

Results. Two new onset detection methods are proposed. They are compared to two estimators taken from the literature review (sAGLR & Komi). One of presented solutions seems to give promising results.

Conclusions. One of presented solutions- Sign Changes algorithm can be widely applied in the area of EMG signal processing. It is more accurate and less parameter-sensitive than three other methods. This estimator can be recommended as a part of ensembled algorithms solution in further development.