Biomechanical Modeling

Muscle force generation: from twitch to tetanus

The mechanical atom of force is the twitch — the brief rise and fall of tension following a single motor-unit action potential, lasting tens to a few hundred milliseconds depending on fibre type. Because the twitch outlasts the action potential that triggered it, successive stimuli arriving before relaxation is complete produce temporal summation: tension builds on residual tension. As firing rate rises, summation passes through unfused tetanus — a rippling plateau in which individual twitches remain visible — and then fused tetanus, a smooth maximal contraction in which the ripples vanish because each twitch begins before the last decays.

The relationship between stimulation rate and steady force is the force–frequency relation, a sigmoid that rises steeply over the physiological firing range and saturates at fusion frequency. Its shape is the mechanical reason rate coding is an effective force control: small increments in firing rate yield large increments in force on the steep portion of the curve. The twitch-to- tetanus ratio — peak twitch force divided by tetanic force — quantifies how much force amplification summation provides, and it differs systematically between fast and slow units.

Recruitment models: size principle meets rate coding

Whole-muscle force is graded by two mechanisms acting together. Recruitment follows Henneman's size principle: motor units are activated in order of increasing size, from small low-threshold (type I, slow, fatigue- resistant) units to large high-threshold (type II, fast, powerful) units, because the same synaptic drive depolarizes small motoneurons first. Rate coding — increasing the firing frequency of already-recruited units — provides the second, continuous lever. The division of labour between them is muscle-specific: small hand muscles complete recruitment at a low fraction of maximal force and rely heavily on rate coding thereafter, whereas large proximal muscles continue recruiting nearly to maximal effort.

A recruitment model formalizes this as a mapping from a common excitatory drive to a set of active units, each with a recruitment threshold and a rate–drive relation; summing the resulting twitch trains, scaled by each unit's force, reconstructs whole-muscle force as a continuous function of neural command. This is precisely why orderly recruitment matters clinically: the same principle that grades force in health is what reduced recruitment (neurogenic) and early recruitment (myopathic) perturb in disease.

Hill-type muscle models: the mechanical substrate

The standard mechanical model of muscle is the Hill-type model, which represents the muscle–tendon unit as three elements. The contractile element (CE) is the active force generator, its output set by activation and by two intrinsic properties. The series elastic element (SEE) — chiefly the tendon and aponeurosis — stores and returns elastic energy and delays the transmission of CE force to the skeleton. The parallel elastic element (PEE) — the passive connective tissue — bears tension as the muscle is stretched beyond its slack length and accounts for passive resistance to elongation.

Two relations make the CE non-linear. The force–length relationdescribes how active force peaks at an optimal sarcomere length where actin–myosin overlap is maximal and falls at shorter or longer lengths. The force–velocity relation(Hill's hyperbola) describes how force declines as shortening velocity increases — a muscle shortening quickly produces less force — and conversely rises above isometric during lengthening. Because muscle force depends jointly on activation, length, and velocity, identical EMG activity at two joint angles or movement speeds corresponds to different forces; the Hill model is what supplies the missing length and velocity dependence that EMG alone cannot.

EMG-driven force estimation and its limits

To convert EMG into force, the signal must first be reduced to an activation estimate. The standard pipeline rectifies the raw signal (taking its absolute value or RMS) and then low-pass filters it to produce a linear envelope, a smooth estimate of the muscle's activation time course. A static or dynamic EMG-to-force mapping — often a non-linear activation function feeding the CE of a calibrated Hill model — then yields predicted force or joint moment. In an EMG-driven model, measured EMG supplies the activation while the musculoskeletal geometry supplies length and moment arm, producing subject-specific force estimates that purely kinematic models cannot.

The limitations are fundamental and must be stated plainly. Cross-talk — pickup from neighbouring muscles — contaminates the surface signal and misattributes activation. The EMG–force relationship is non-linear and depends on length and velocity, so a single calibration rarely holds across the range of motion. Fatigue dissociates EMG from force entirely: as the spectrum compresses and amplitude drifts, the same EMG corresponds to declining force, breaking any fixed mapping. And EMG captures only the net agonist drive, blind to antagonist co-contraction that alters the joint moment without changing the recorded agonist signal. These constraints make EMG-driven force a powerful estimate but never a measurement.

Functional-output prediction: prosthetics and rehabilitation

The payoff of these models is the prediction of functional output — translating recorded muscle activity into intended movement or force for an external device. In myoelectric prosthetics, surface EMG (increasingly decomposed into motor-unit firings via high-density grids) is mapped to the intended grasp, force, or joint angle, letting an amputee control a limb by the residual muscle's neural command. In rehabilitation and exoskeletons, EMG-driven estimates of joint moment drive assistive torque in proportion to the user's own effort, and serve as biofeedback to retrain activation patterns. The accuracy of every such system is bounded by the same modeling chain — recruitment, the force–frequency and force–length–velocity relations, and the EMG-to-force mapping — so the physiology in this lesson is literally the transfer function that determines whether the device feels natural or fails the user.