Page 318 - Biomedical Engineering and Design Handbook Volume 1, Fundamentals

ELECTROMYOGRAPHY AS A TOOL TO ESTIMATE MUSCLE FORCES 295

usually preserve the important contents in the EMG signal. Only the information relative to the firing
rates of active motor units, in the range of 5 to 25 Hz for most muscles (Enoka, 2002), is disregarded.
However, this information is not of great importance for most applications.
After the EMG is high-pass-filtered, the processed signal still has many sharp peaks. The power
spectrum from surface EMG signal shows that the majority of the signal power is at frequencies
below 350 Hz (Wakeling and Rozitis, 2004). The rest of the high-frequency contents are mostly noises,
which may be generated by surrounding electromagnetic devices, and poor electrode-skin contact
(Cram and Garber, 1986). A low-pass filter can then be applied to eliminate the high-frequency noises.
The cutoff frequency is usually chosen close to 500 Hz (De Luca, 1997). The filter should also have
zero-phase delay properties (e.g., forward and reverse pass fourth-order Butterworth filter).
If the EMG signal will be used to estimate muscle force, more high-frequency contents of the
signal should be eliminated, because muscle naturally acts as a filter and we want this to be charac-
terized in the EMG-force transformation. That is, although the electrical signal that passes through
the muscle has frequency components over 100 Hz, the force that the muscle generates is of much
lower frequencies (e.g., muscle force profiles are smoother than raw EMG profiles). In muscles there
are many mechanisms that cause this filtering; for example, calcium dynamics, finite amount of time
for transmission of muscle action potentials along the muscle, and muscle and tendon viscoelasticity.
Thus, in order for the EMG signal to be correlated with the muscle force, it is important to filter out
the high-frequency components with a lower cutoff frequency. The range of 3 to 10 Hz is typical for
the cutoff frequency in the EMG-force transformation (Buchanan et al., 2004).

12.3.3 Different Approaches to Analyze EMG
After the unwanted noises are removed from the EMG signal, different approaches may then be used
to analyze the signal.
Three parameters are commonly calculated in the time domain to provide useful measurement of
EMG signal amplitude: root-mean-squared (RMS) value, mean integrated rectified value, and average
rectified value (De Luca and Vandyk, 1975). RMS value represents the signal power during voluntary
contractions, and thus is recommended above the other two parameters (De Luca, 1997).
The power density spectrum of EMG signal is commonly used to analyze the EMG in the
frequency domain. A fast Fourier transform can be used to obtain the power density spectrum of the
signal. Mean frequency, median frequency, and bandwidth of the spectrum are the three main para-
meters used to provide useful measurement of the spectral properties of the signal. The power spec-
trum of EMG signal is typically a skewed bell-shaped curve, and the majority of the signal power lies
between 50 to 150 Hz (Wakeling and Rozitis, 2004). The power spectrum of the EMG signal will be
changed during the progression of muscle fatigue, and this spectrum modification property can be used
in the study of muscle fatigue. The median frequency of the power spectrum is often used as a fatigue
index as a measure of muscle fatigue.
All of the above approaches are implemented to analyze the contraction of a muscle. Meanwhile,
the EMG signal can also be decomposed into individual MUAPTs, which can help researchers to
study motor unit properties and behavior (Stashuk, 2001).

12.4 EMG-DRIVEN MODELS TO ESTIMATE MUSCLE FORCES

EMG-driven models use EMG and joint kinematics, recorded during a range of static and dynamic
trials, as input to estimate individual muscle forces and joint moments (Lloyd and Besier, 2003;
Buchanan et al., 2004; Buchanan et al., 2005). Given appropriate anatomical and physiological data,
these models can be applied to any joint, such as the elbow joint (Manal et al., 2002), the knee joint
(Lloyd and Besier, 2003), and the ankle joint (Buchanan et al., 2005).
This overall model is composed of four main parts: (1) anatomical model, (2) muscle activation
dynamics model, (3) muscle contraction dynamics model, and (4) model of calibration process. Here
we will discuss the details involved in each of these steps for an ankle and knee model as an example.