**8.1. Input range and resolution of the A/D converter**

The input range is a parameter associated with resolution and indicates the range of voltage that the A/D converter board can represent numerically. This band can be 5 V, 2.5 V, 0 to 5V, 10V etc.

When the input signals do not fall within the A/D card's range, it is necessary to condition them (amplify or attenuate) before inputting them into the A/D converter. Figure 6 shows an example in which the A/D converter or the conditioning gain is misaligned with the signal. Figure 7 depicts a gain adequate for visualizing the EMG signal.

**Figure 6.** A/D converter range at odds with the amplification gain [3].

**Figure 7.** Properly aligned A/D converter range and amplification gain [3].

The resolution of an A/D converter indicates the lowest variation in analog signal that the converter can detect, which is generally presented in bits. Thus, converter resolutions can be 10, 12, 14 or 16 bits, etc., with the most common being 12- and 16-bit.

A converter with a 5V input range and a resolution of 12 bits can represent the input signal in 4096 (212) divisions and levels or detect changes of 2.4 mV (10 V divided by 4096 levels). A 16-bit converter may represent the same signal in 65536 (216) divisions and detect changes at levels of 153 μV. (10 V divided by 65,536 levels), [4].

#### **8.2. Sampling rate**

Computational Intelligence in Electromyography Analysis – 400 A Perspective on Current Applications and Future Challenges

**8.1. Input range and resolution of the A/D converter** 

Figure 7 depicts a gain adequate for visualizing the EMG signal.

**Figure 6.** A/D converter range at odds with the amplification gain [3].

An analog-digital (A/D) converter converts analog signals (EMG goniometry, force transducer) into digital data. The digitized signal can then be processed by the computer.

The input range is a parameter associated with resolution and indicates the range of voltage that the A/D converter board can represent numerically. This band can be 5 V, 2.5 V, 0 to

When the input signals do not fall within the A/D card's range, it is necessary to condition them (amplify or attenuate) before inputting them into the A/D converter. Figure 6 shows an example in which the A/D converter or the conditioning gain is misaligned with the signal.

**8. Analog-digital converter** 

5V, 10V etc.

In practice, the input signal to the A/D converter varies over time; the goal is to record this variation. Since a computer's storage capacity is finite, the recording can only continue for a limited time.

The discretization of time is carried out by sampling the signal at regular intervals. The reverse of this interval is the sampling rate. For example, at a sampling rate of 100 samples per second (i.e., 100 Hz), the interval between samples is 10 ms. The sampling rate is equivalent to the resolution of the A/D conversion but applied to time.

However, due to the limited space available for data storage, there is a compromise between the sampling rate and the duration of acquisition. For example, for sampling rate of 100 samples per second, the maximum acquisition will be 166 minutes and 40 seconds. By increasing the rate to 1000 samples per second, the maximum is 16 minutes and 40 seconds.

The sampling rate must also be very low compared to the frequency of signal variation due to the effects of sub-sampling (aliasing).

Application of Surface Electromyography in the Dynamics of Human Movement 403

Correcting the curve is an operation normally used to enable the subsequent integration of the signal, since it transforms a curve containing both positive and negative values (Figure

There are two ways to rectify the curve: eliminating the negative values (half-wave rectification), or reversing the negative values and adding them to the positive values (full wave rectification). Full-wave rectification has the advantage of maintaining all of the

The RMS is the amount of continuous signal able to contain the same amount of energy. It is mathematically defined as the square root of the mean of the squares of the instantaneous

One problem when comparing different EMG signals has to do with differences in the

Normalizing means transforming, without changing the signal's structure, the duration differences into signals with the same number of samples. This can be done, for example, by taking the signal containing the lowest number of samples as a reference. An algorithm can be applied that determines, depending on the duration of each signal, the number of samples to be removed at certain intervals, reducing all signals to the same number of samples contained in the shorter of the two signals, and thus retaining the

The EMG signal varies greatly upon comparison with recordings from the same individual or different individuals. The absolute value of the EMG signal thus provides little information, especially when dealing with signals from different individuals or the same individual at different times. One way to compensate for this limitation is to normaliz EMG amplitude curves. This technique consists of transforming the absolute amplitude values of the different

The mathematical interpretation of the integral concept consists of determining the area enclosed by curve, whether an EMG or any other signal. In the case of the EMG, so that the result of integration is not zero, a rectified signal must be used. By integrating the EMG signal, a result that is proportional to the number of electrical impulses is obtained [3].

curves to be compared into values relative to a reference EMG taken as 100% [4, 7, 15].

10) and a zero mean to a curve of only positive absolute values (Figure 11).

information contained in the signal, unlike half-wave rectification [5, 28].

*9.1.2. Signal rectification* 

*9.1.3. Root-mean-square value of the signal* 

*9.1.4. Normalization of the signal in the time domain* 

duration of the various signals to be compared.

values of the signal [4, 12, 22, 23].

original forms [16].

*9.1.5. Amplitude normalization* 

*9.1.6. Integral of the EMG signal* 

An aliasing effect occurs whenever the sampling frequency is less than twice the highest frequency component of signal frequency, according to the Nyquist theorem [12].

EMG recording is usually done at a maximum frequency of 500 Hz, and the sample should be at least 1000 Hz. To analyze muscle activity in the most comprehensive way possible, it is advisable to work with a sampling rate on the order of 2000 Hz, with the highest frequency component of the signal always limited by the low-pass filter [4, 12, 28].
