Biosignal Preprocessing

 

Following transduction at the tissue/sensor interface, most biosignals are still in analog (continuous) form. The signals are then processed and stored either in analog form or at some point they may be converted to digital. Analog computers and processors are those that operate on continuous signals. Examples of electronic devices that work with analog signals include: operational amplifiers (amplify the signal), filters which use discrete electrical components such as resistors and capacitors, and storage media such as a phonograph recording device. The main advantage to working with biosignals in the analog form is that the entire continuous signal is recorded. If the independent variable of the biosignal is time (usually the case) then no time information will be lost. However, analog components can be difficult to work with in the sense that any change in amplification, filtering, or other mathematical operation that is imparted to the signal must usually be done manually by physically changing the electrical elements of the circuit. Storage of continuous signals can require a significant amount of physical space. In many cases, it is advantageous to digitize the biosignal at some point prior to storage of the data. The reason is that it is often simpler to process and store, as we shall see.

Both analog and digital signals often require some sort of preprocessing steps to ensure that the data is stored in a form that is of high quality. Firstly, the signal must be amplified. This should be done at a point in the electronic circuit that is as close to the transduction process as possible, to minimize the amount of noise added to the biosignal. Typically, bioelectric signals are on the order of millivolts in size. Since analog-to-digital conversion integrated circuits (IC’s) typically require inputs in the range of 1-10 volts, bioelectric signals are often scaled by 100x to 1000x. Operational amplifiers set to a gain of 100x – 1000x are used for this purpose. A discrete component can be used to amplify the signal, or an IC can be used. During or immediately after the amplification process, the biosignals are usually band-pass filtered. This means that both the low and the high frequencies are removed, and intermediate frequencies of the biosignal are passed through the filter. A band-pass filter can also be thought of as a combination of a low and a high pass filter. The high pass filter removes the low frequencies (smooth components) of the signal. Such filters ALWAYS remove the average level of the signal (DC component). It is often but not always desirable to remove the DC component since offsets from the 0 volt level, in the case of bioelectric signals, are often caused by noise (DC bias). Low pass filters remove the high frequency components. These low pass filters are also called "anti-aliasing" filters. Their purpose is to remove high frequency noise from the signal prior to any digitization. If high frequency noise is not removed, and if the signal is digitized at less than twice the frequency of the highest significant noise component, aliasing will occur. This means that the signal will be distorted because the number of digital points used to represent it is insufficient to describe the high frequency components.

The INA102 integrated circuit, manufactured by the Burr Brown company, serves as what is called a "first stage" amplifier. This means that this is the first amplifier seen by the signal. The chip has 16 pins, and comes in a standard DIP (dual in-line package). Two of the pins on the chip serve as inputs, since the INA102 is a differential amplifier. There are also pins which determine the gain. Depending on how these pins are configured, the gain will be "hard-wired" to 10x, 100x, or 1000x. We can set the gain to 100x, for example, and further amplify the signal using a second stage amplifier. There are also pins on the chip for ± power, and ground. Finally, the chip has a single output pin. There is an ample supply current available at the output to drive subsequent electronic components in the circuit (it acts as a buffer). What this means is that the voltage level of the output signal will remain stable even if the next chip in the circuit has a low input impedance and sinks a lot of current. This is important so that the signal does not become distorted.

A series of "passive" resistors and capacitors can then be used to filter the signal. For a simple high or low pass filter, only 1 resistor and 1 capacitor is needed in the circuit. For a high pass filter, the capacitor is placed in series, so that the biosignal must pass through it, and the resistor is "tied" to the line after the capacitor, with the other end of the resistor connected to ground. We can describe the transfer function for the high pass filter according to the current passing through both impedance elements as approximately:

(Vout – Vin)/ (jw C)-1 = 0 - Vout / R

Vout – Vin = Vout / (jw RC)

Vout/Vin = jw RC / (1 + jw RC)

 

Where w = 2p f and f is the frequency in cycles per second (Hz). This is a high pass filter. When the value of f, and therefore of w , is low, the value of the above equation approaches zero, i.e., the output approaches zero volts no matter what the values of the input is. Whereas, as high frequencies, the 1 in the denominator becomes negligible, and the entire fraction approaches a value of one (unity); hence, the output is approximately the same as the input.

 

By comparison, a low pass filter looks like this:

(Vout – Vin)/R = 0 - Vout / (jw C)-1

Vout (1 – jw RC) = Vin

Vout/Vin = 1/(1 + jw RC)

 

This is just the opposite of the high pass filter. When the value of f, and therefore of w , is low, the denominator approaches a value of one and therefore the value of the above equation approaches one (unity); hence, the output is approximately the same as the input. Whereas, at high frequencies, the term in the denominator becomes large, and the entire fraction approaches a value of zero, i.e., the output approaches zero volts no matter what the values of the input.

Often, anti-alias filters in the form of an integrated circuit are used to low pass filter the signal. Although ICs are active circuits (they require power from a power supply), anti-alias ICs are typically better at low pass filtering than those produced with passive components. The reason is that the sophisticated circuitry within the IC produces a very sharp "cutoff" the point in the frequency domain at which the signal is not passed through the filter. An example is the LTC1062 IC from manufactured by the Linear Technology company. This IC has an input for the signal, and pins to change the cutoff frequency of the low pass filter. It also has pins for power and ground, and an output. The cutoff frequency of the low pass filter can be adjusted by the software of the computer using another chip called a "switch". An output port from the computer can be used to control the switch. If the level of the switch is a binary 1 (on) the anti-alias filter will have a cutoff frequency of 1000Hz. If the level of the switch is a binary 0 (off) the anti-alias filter will have a cutoff frequency of 500Hz. Since the anti-alias characteristic is software selectable, it is easy to change during the experiment or for different experiments depending on the characteristics of the biosignals.