Thefirst recorded EEG frequencies were alpha rhythms of about 10 Hz, which wererecorded by Berger (1929) and later in the beta. The beta value is now used todefine the frequency range of about 12 Hz to 30 Hz. The range of Gammafrequency oscillations found by Adrian (1942) is between 30 and 80 Hz.
Thefrequency ranges are not precisely defined, and the intervals are useddifferently in different articles 28.The purpose of signal processing methods is toacquire knowledge, to make it understandable and to interpret it for this usingEEG signals may be appropriate. Our brain gives different reactions atdifferent times, but our first target is the reactions that the brain givesabout the heart and the effects that those reactions have on the signals. Thebrain signals will be used for this purpose to examine the functioning of thebrain with the best and most understandable signals produced, processed and interpreted as it is possible. Thismethod is called brain signal processing.
29Normally there are many things or data that donot work in the obtained signals and when the signals are received, the signalis distorted due to some negative factors in the environment. So, by filteringthe signal, we can get signal processing at the desired frequency range. Forvarious purposes, brain functioning is based on well-processed signals 30.In Figure 6, an EEG signal of a schizophreniapatient which plotted in MATLAB.EEGrecords have been obtained from patients such as epilepsy, schizophrenia, etc.
,through the appropriate registration conditions prepared by the experts of theDepartment of Neurology. Since EEG signals below 3 Hz and above 40 Hz areunreliable due to the low signal-to-noise ratio, all of EEG data is bandpassfiltered to between 3-40 Hz 31. Numerical filters are one of the methods that convert inputsignals into desired output signals.
Separating the cluttered signals, thenoises in the signals and increasing the quality of the signal are some of thepurposes of this filter (Figure 7). Digital filters, usually identified by the before mentioneddifferences, allow the components of the determined signal frequencies andtheir coefficients to be selected correctly 32. Finite Impulse Response (FIR) andInfinite Impulse Response (IIR) are two main digital filters commonly used indigital signal processing 33. The main purpose of the filters to be designed in this workis to reduce the noise in the EEG signal. For this purpose, it will be appliedto brain signals with FIR (Finite Impulse Response) and IIR (Infinite PulseResponse) filter designs to be designed in MATLAB program. The EEG signaloperating frequency range is 0.
5 to 38 Hz. For this reason, the cutofffrequencies of the FIR and IIR filters are designed to be 0.5 to 45 Hz todesign the filter.
A bandpass filter was used in the design. For filter design,MATLAB Filter Design and Analysis Department which is run by”fdatool” command is used. Once the designed filters are applied tothe EEG signal, the characteristics are removed from the filtered EEG signal 34.The Parks-McClellan algorithm is used to design the mostappropriate filters between the real and desired frequency responses. Filtersare ideal for the reduction of the maximum error between the desired frequencyresponse and the actual frequency response.
The filters designed in this wayexhibit an equal behavior in the frequency response and are sometimes referredto as equiripple filters 37.Least Squares Thelinear-phase FIR filter is used to minimize the weighted, integrated squareerror between the ideal partial linear function and the magnitude response offilter the preferred frequency bands 38.Designing FIR filters using the windowing method can besimple and fast. One of the simplest window methods is the rectangular windowmethod 39.
Constrained Least Squares (CLS) FIR filter design functionsare one of the techniques that enable FIR filters to be designed without havingto clearly define passbands for magnitude response 40.Filter designs were made using some of these methods, and itwas applied to signal.IIR filters can be the best efficient of the digital signalprocessing(DSP) methods 41. As the inputis bound to its current and past values, which is also connected to the pastvalues, that is feedback filtered.
The relationship between inputs and outputsof an IIR Filter is showing in the Equation 2.A Butterworth filter is a type of signal processing filterthat has a straight frequency response as possible as within the passband 42.