A transient change in electrical potential across the axonal or muscle fibre membrane, which excites adjacent sections to propagate the change in potential along the nerve or muscle fibre. Action potentials can also be generated in other types of cells with excitable membranes. The term is also used to refer to the net potential change recorded during the simultaneous discharge of more than one muscle fibre, such as the motor unit action potential and the compound muscle action potential.

The difference between the value of the analog signal and the closest available discrete value, at each sampling instant, provided by the analogue to digital (A/D) converter. Related term: A/D conversion, A/D resolution. It is reduced by increasing the number of levels of the A/D converter which is given by 2ⁿ where n is the number of bits, e.g., a 4-bit converter has 16 levels (2⁴), from 0 to 15.

The smallest incremental change in voltage in a continuous analog signal that will change the discrete binary value detected by an A/D converter by one count. Related term: A/D conversion, A/D quantization error. An n-bit A/D converter produces 2ⁿ discrete binary levels. The resolution is the input voltage range of the A/D converter divided by 2ⁿ⁻¹ steps.

Muscle contributing to the force and/or joint moment required to achieve the task goal, often called a “prime mover”. Antagonist and agonist muscles often occur in pairs (antagonistic pairs). In anatomy, the role of certain muscles is often defined based solely on their anatomical location. However, this idealized terminology is less applicable in the field of biomechanics, where it is more difficult to ascribe a specific role to a muscle during complex multi-joint motor tasks.

The distortion of a signal and its power spectrum that occurs when the analog signal is sampled at a rate that is too low to create an accurate digital representation of the signal. Related term: Sampling in time. The sampling rate should be greater than twice the highest frequency component contained in the signal to be recorded to avoid aliasing (Nyquist’s theorem). In practice, the sampling rate for recording EMG signals should be at least 2.5–3 times higher than the highest frequency component contained within the EMG signal. To avoid aliasing, signals should be low-pass filtered with an antialiasing filter with a cut-off frequency less than half the desired sampling frequency before sampling (Nyquist’s theorem).

The procedure by which a measurement obtained from a particular muscle and subject is expressed as a percentage or proportion of a reference measurement obtained from the same muscle and subject. For example, the amplitude of an EMG signal might be expressed as a fraction of the EMG signal amplitude recorded during an MVC of the same subject.Normalization of the EMG amplitude to a reference value enables comparisons between subjects/muscles/measurement sessions/ electrode positions, etc. To interpret and compare the RMS or ARV amplitude of EMG signals they must be normalized, as a number of factors that are unrelated to muscle activity can influence the recorded voltage (e.g., thickness of subcutaneous layers).

The process of converting a continuous analog signal into a sequence of discrete (binary) numbers that can be stored and processed. Related terms: Sampling in time, aliasing, A/D quantization error. Choosing an A/D converter with a higher number of bits can reduce quantisation errors (the most commonly used A/D converters have resolutions of 8, 12 or 16 bits). The number of bits in an A/D converter determines the number of discrete voltage levels that can be represented by the A/D converter. For example, a 16-bit A/D converter can represent 2¹⁶ = 65536 (0 to 65535) distinct voltage levels (i.e., 2ⁿ where n is the number of bits).

Muscle that produces force and/or joint moment in a direction that opposes the action of the agonist muscle. Agonist and antagonist muscles may be activated together to increase stability in the joint (i.e., increase joint mechanical impedance) to better control a specific task. It is good practice to monitor both.

A measure of self-similarity within a signal, i.e., the correlation between the signal and a delayed version of itself. In other words, how much the value of a signal depends on (or is related to) its past and future values. The discrete auto-correlation function, R_xx, of a finite, discrete signal, x, can be estimated as in Formula 1. N is the total number of samples in the signal, m is the lag in samples (i.e., |m| = 0, 1, 2, … M⁻¹, M is the range of lag values where M ≪ N), and x_n is the value of the n^th sample within the signal. Since R_xx[−m] = R_xx[m], it is necessary only to evaluate the auto-correlation function for the nonnegative values of m. This is a biased estimate of R_xx, but the bias is small where M ≪ N. The equation for R_xx in continuous time can also be expressed as (note that below we provide the expression over the entire time axis):

The range of frequencies in which a signal contains power and information. The 3-dB bandwidth is the interval between frequencies at which the power spectral density has dropped to half power, or 3 dB below the peak value. Surface EMG signals typically contain energy at frequencies from 10 to 500 Hz. Intramuscular EMG signals contain energy at frequencies as high as 5 kHz (due to the rising edges of action potentials from muscle fibers close to the electrode). The bandwidth of the recorded signal may be altered by hardware or software filters applied to the signal.