METHOD FOR OBTAINING AN INDICATOR OF PRESENCE FOR ALZHEIMER'S DISEASE USING EEG SIGNALS TECHNICAL FIELD [0001] The present disclosure relates to a method for obtaining an indicator of presence for Alzheimer's disease using EEG signals, namely by training an ANN. BACKGROUND [0002] In Rodrigues et al [21], it is disclosed a multiband cepstral and lacstral analysis via a Discrete Wavelet Transform (DWT) applied to each EEG segment. A Kruskal-Wallis test was applied for feature selection. A set of machine learning tools were fed with the features combination provided by a genetic algorithm for classification. [0003] The document EP3241489 discloses predictive neuromarkers of Alzheimer's disease, which consist of at least one spectral feature acquired from EEG signals of an individual, as well as at least one Riemannian distance between a spatiofrequential covariance matrix generated from the EEG signals of the individual and one or more reference spatiofrequential covariance matrices. Additionally, the invention involves a non-invasive technique for detecting Alzheimer's disease in an individual by utilizing the predictive neuromarkers of the disease. [0004] The document US2019200893 discloses methods and systems to facilitate the analysis and evaluation of complex, quasi-periodic system by generating computed phase-space tomographic images as a representation of the dynamics of the quasi-periodic cardiac systems. The computed phase-space tomographic images can be presented to a physician to assist in the assessment of presence or non-presence of disease. In some embodiments, the phase space tomographic images are used as input to a trained neural network classifier configured to assess for presence or non-presence of significant coronary artery disease. [0005] The document EP2584963 discloses a method for analysis of the extent of conscious awareness and likelihood of recovery of a patient includes the steps of applying to the patient a sensory stimulus sequence which is typically auditory; carrying out an EEG to generate waveform signals to record changes in the electromagnetic fields generated by the patient's neural activity; using software provided in a processor to process the waveform signals in order to locate waveform peaks, identify the event-related potential (ERP) components obtained in the waveform and to obtain quantitative measures of those components; and using the software to generate and communicate scores indicative of the extent of conscious awareness and likelihood of recovery of the patient. [0006] These facts are disclosed in order to illustrate the technical problem addressed by the present disclosure. GENERAL DESCRIPTION [0007] The present document discloses a method for obtaining an indicator of presence for Alzheimer's disease using electroencephalogram (EEG) signals, herein named Margolacs, to characterize the Alzheimer’s disease (AD) activity and evolution prospecting for different AD stages: Mild Cognitive Impairment (MCI), Mild and Moderate AD (ADM) and Advanced AD (ADA). [0008] A set of features extracted from EEG signals, processed with the disclosed method, feed an Artificial Neural Networks (ANN) with aim to diagnose and prospect AD in an automatic way. The disclosed method allowed the ANN to reach an accuracy of 98 % for All vs All, 99 % for healthy control group (HC) vs MCI, 99 % for HC vs ADM, 97 % for MCI vs ADM- ADA. [0009] In HC vs MCI, HC vs ADM and MCI vs ADM-ADA, the method of the present disclosure outperforms the state-of-art methods by 1 %, 3 %, and 3 %, respectively. In All vs All, it outperforms the state-of-art EEG and non-EEG methods by 2 % and 4 %, respectively. These results indicate that the proposed method represents an improvement in diagnosing AD. [0010] It is further disclosed an interface that provides to users, e.g., doctors, important information for diagnosis, Alzheimer Scalp-level activity prospection, Power spectral density metrics, recording data and the corresponding EEG conventional sub-bands plot. [0011] A method for obtaining an indicator of presence for Alzheimer's disease using EEG signals from a subject, comprising the steps: obtaining an EEG signal of a subject; applying an analysis filter bank for processing the obtained EEG signal into a plurality, m, of subband signals, x _m ( n); applying the module to each subband signal, to obtain the module of each subband signal, |x _m ( n)|; applying a logarithm to each subband signal's module, |x _m ( n)|, to obtain the logarithmic magnitude of each subband signal; applying synthesis filter bank to each subband signal for reconstructing a corresponding subband signal, y _m ( n); applying the module to the reconstructed subband signal, y _m ( n), to obtain an output subband signal,x ^l ( n); calculating a set of distance functions for obtaining distances between output subband signals, x ^l ( n); feeding the calculated set of distances to a pre-trained machine learning classifier to obtain a classification score; wherein the classification score is indicative for the Alzheimer's disease presence. [0012] It is disclosed a method for obtaining an indicator of presence for Alzheimer's disease using EEG signals from a subject, comprising the steps: obtaining an EEG signal of a subject; applying an analysis filter bank for decomposing the obtained EEG signal into a plurality, m, of subband signals, x _m ( n); applying a module to each subband signal, to obtain a module of each subband signal, |x _m (n)|; applying a logarithm to each subband signal's module, |x _m (n)|, to obtain a logarithmic magnitude (i.e. the logarithmic of the module of each subband signal) of each subband signal; applying a synthesis filter bank to each subband signal logarithmic magnitude for reconstructing a corresponding subband signal, y _m ( n); applying a module to each reconstructed subband signal, y _m ( n), to obtain an output subband signal, x ^l (n); calculating a set of distance functions for obtaining distances between output subband signals, x ^l (n); feeding the calculated set of distances to a pre-trained machine learning classifier to obtain a classification score; wherein the classification score is indicative for the presence of Alzheimer's disease. [0013] In an embodiment, the analysis filter bank is a discrete wavelet transform, DWT. [0014] In an embodiment, the synthesis filter bank is an inverse discrete wavelet transform, inverse DWT. [0015] In an embodiment, the classification score comprises a plurality of classes corresponding to an indicator of a plurality of Alzheimer’s disease stages and to a stage of non-Alzheimer’s. [0016] In an embodiment, the plurality of Alzheimer’s disease (AD) stages comprises Mild Cognitive Impairment, Mild and Moderate AD, and Advanced AD. [0017] In an embodiment, the pre-trained machine learning classifier comprises an output layer with a node for each one of the classes. [0018] In an embodiment, the number of subband signals is five. [0019] In an embodiment, the subbands comprise a δ-band for 1-4 Hz, a θ-band for 4-8 Hz,α-band for 8-13 Hz, ^^-band for 13-30 Hz and a γ-band for 30-40 Hz. [0020] In an embodiment, the set of distance functions comprises a normalized root mean square distance, defined as [0021] In an embodiment, the set of distance functions comprises a normalized root mean square distance, defined as [0022] In an embodiment, the set of distance functions comprises a Euclidean distance, defined as [0023] In an embodiment, the set of distance functions comprises a quefrency weighted distance, defined as [0024] In an embodiment, the set of distance functions comprises a quefrency rooted weighted distance, defined as [0025] In an embodiment, the set of distance functions comprises a quefrency squared weighted distance, defined as [0026] In an embodiment, the classification score comprises a plurality of indicators in a 2D map for each EEG channel. [0027] It is also disclosed a method comprising calculating: where y _m ( n) is and x _m (n) is a mth subband signal, for obtaining calculated features for feeding to a pre- trained machine learning classifier. [0028] It is also disclosed a method for obtaining a scalp-level activity 2D map using a plurality of EEG signals from a subject, comprising the steps: for each of the plurality of EEG signals, applying the method for obtaining an indicator of presence for Alzheimer's disease using EEG signals from a subject according to any of the disclosed embodiments; generating a scalp-level activity 2D map with the obtained classification scores for the plurality of EEG signals; wherein the scalp-level activity is indicative for the Alzheimer's disease presence. [0029] It is also disclosed a device for obtaining an indicator of presence for Alzheimer's disease by EEG signals comprising an electronic data processor arranged to carry out the method according to any of the disclosed embodiments. [0030] It is also disclosed a method for training a machine-learning model for obtaining an indicator of presence for Alzheimer's disease using EEG signals from a plurality of subjects, comprising the steps: obtaining EEG signals from the plurality of subjects; applying an analysis filter bank for decomposing the obtained EEG signals into a plurality, m, of subband signals, x _m (n) for each subject; applying a module to each subband signal, to obtain a module of each subband signal, |x _m ( n)|; applying a logarithm to each subband signal's module, |x _m ( n)|, to obtain a logarithmic magnitude of each subband signal; applying a synthesis filter bank to each subband signal logarithmic magnitude for reconstructing a corresponding subband signal, y _m ( n); applying a module to each reconstructed subband signal, y _m ( n), to obtain an output subband signal, x ^l (n); calculating a set of distance functions for obtaining distances between output subband signals, x ^l (n) for each subject; feeding the calculated set of distances to a pre-trained machine learning classifier to obtain a classification score for each subject; calculating a classification error between the calculated classification score and a previously obtained classification for each subject; updating the machine-learning model using the calculated classification error; wherein the classification score is indicative for the presence of Alzheimer's disease. [0031] In an embodiment, the obtained EEG signals are one EEG signal per subject. [0032] In an embodiment, the obtained EEG signals are a plurality of EEG signals per subject. [0033] It is also disclosed a method comprising the steps for updating the machine-learning model: applying a feature selection to the calculated set of distances to obtain a feature vector; applying a feature normalization to the feature vector to obtain a normalized feature vector; applying a leave-one-out cross-validation to the normalized feature vector, wherein the cross-validation is applied to a feature vector of a different subject for each EEG channel. [0034] It is also disclosed a device for training a machine-learning model for obtaining an indicator of presence for Alzheimer's disease using EEG signals according to the training method of any of the disclosed embodiments. [0035] It is also disclosed a system for obtaining an indicator of presence for Alzheimer's disease by EEG signals comprising: a set of EEG sensors; a device configured to carry out the method of any of the disclosed embodiments; wherein the set of EEG sensors is arranged on a support for placing on a head of a subject. [0036] In an embodiment, the system comprises a display configured to display the EEG signal and calculated classification score. BRIEF DESCRIPTION OF THE DRAWINGS [0037] The following figures provide preferred embodiments for illustrating the disclosure and should not be seen as limiting the scope of invention. [0038] Figure 1: Schematic representation of an embodiment of the data processing for training an ANN. [0039] Figure 2: Schematic representation of an embodiment of the method for obtaining an indicator of presence for Alzheimer's disease using EEG signals. [0040] Figure 3: Schematic representation of an embodiment of disclosure. [0041] Figure 4: Schematic representation of an embodiment of an interface. DETAILED DESCRIPTION [0042] A method for obtaining an indicator of presence for Alzheimer's disease using EEG signals from a subject, comprising the steps: obtaining an EEG signal of a subject; applying an analysis filter bank for processing the obtained EEG signal into a plurality, m, of subband signals, x _m (n); applying the module to each subband signal, to obtain the module of each subband signal, |x _m ( n)|; applying a logarithm to each subband signal's module, |x _m ( n)|, to obtain the logarithmic magnitude of each subband signal; applying synthesis filter bank to each subband signal for reconstructing a corresponding subband signal, y _m (n); applying the module to the reconstructed subband signal, y _m ( n), to obtain an output subband signal,x ^l (n); calculating a set of distance functions for obtaining distances between output subband signals, x ^l (n); feeding the calculated set of distances to a pre-trained machine learning classifier to obtain a classification score; wherein the classification score is indicative for the Alzheimer's disease presence. [0043] Figure 1 shows a schematic representation of an embodiment of the data processing for training an ANN, where 101 represents a database, 103 represents calculated distances, 105 represents feature extraction, 107 represents feature normalization, and 109 represents an ANN. [0044] Figure 2 shows a schematic representation of an embodiment of the method for obtaining an indicator of presence for Alzheimer's disease using EEG signals, where 102 represents a represents EEG data, 103 represents the disclosed method, 105 represents the features extraction, 107 represents the features normalization, 110 represents a trained and validated ANN, and 111 represents an indicator of presence for Alzheimer's disease. [0045] The dataset contains EEG recordings from 38 subjects: 11 healthy subjects as Control group (C), 8 patients with MCI, 11 patients with ADM and 8 patients with ADA. The patients were not getting any medication during the signals’ acquisitions, and they were tracked for four years by the neurological department at the University Hospital Centre of São João, where their diagnoses have been validated. The average age and average MMSE score of the participants in each study group are presented in Table 1. All participants gave their consent prior to participating in this study, whose protocol number CES198-14 was approved and authorized on 20 March 2015 by the local Ethics Committee of the University Hospital Centre of São João (UHCSJ), Porto, Portugal. [0046] Table 1: Information about the EEG dataset. [0047] The EEG samples were recorded from the 19 scalp loci of the International 10-20 configuration using a digital electroencephalograph in UHCSJ, Porto, Portugal. The participants were in a state of relaxation and with eyes closed. The sampling frequency was 256 Hz. The recordings were filtered through a digital band-pass filter with cut-off frequencies of 1 and 40 Hz, thus passing the frequency range of the conventional EEG subbands. The EEG channel signals were split into non-overlapped artifacts-free segments of a pre-defined period, e.g., 5s-long segments, 1280 samples, as done in [18-20]. [0048] The amplitude of each signal was normalized according to the equation: where N represents the signal size, x _input (n) represents EEG input data, and the signal mean value was removed. [0049] The discrete wavelet transform (DWT) of a discrete-time finite-energy signal is its decomposition in a set of basis functions obtained from a finite number of prototype sequences and their time shifted versions [23]. It is an optimal tool for time-frequency signal analysis [23]. [0050] This structured expansion and its corresponding reconstruction are implemented by means of an octave-band critically decimated filter bank [12, 23]. Considering only the positive frequencies, the mth subband is confined to where S + 1 is the number of subbands and π is the normalized angular frequency. [0051] The DWT uses a synthesis scale function n n Φ ₁ ( ) and a synthesis wavelet function ψ ₁ ( ) defined as [12] and n n where g _LP ( ) and g _hP ( ) are the impulse responses of the half-band low-pass and high-pass synthesis filters, respectively. [0052] Defining the following recursion formulas [12] where the symbol ∗ denotes the convolution operator, the equivalent synthesis filter of the mth subband is given by [12] [0053] The synthesis subband filters g ₁ (n) to g _s (n) are roughly dilated versions of the wavelet ψ ₁ (n), thereby having approximately constant shapes, and g ₀ (n) is a roughly dilated version of the scaling function Φ ₁ (n), which is the basic low-pass filter of each reconstruction stage [12]. [0054] The DWT uses an analysis scale function and an analysis wavelet function defined as and where h _LP (n) and h _HP (n) are the impulse responses of the half-band low-pass and high-pass analysis filters, respectively. [0055] Defining the following recursion formulas the equivalent analysis filter of the mth sub-band is given by [0056] The analysis subband filters h ₁ (n) to h _s (n) are roughly dilated versions of thereby having approximately constant shapes, and h ₀ (n) is a roughly dilated version of which is the basic low-pass filter of each decomposition stage. [0057] If the wavelet base is biorthogonal, then the filter bank performs a perfect signal reconstruction [12,23]. If the wavelet base is orthogonal, then the filter bank is of a perfect reconstruction type and the analysis filters are identical to the time-reversed synthesis filters [12,23], that is, where L _m is the length of the mth subband filter. [0058] The mth subband signal, m = 0, 1, … , S, is given by [0059] In an embodiment, the DWT is applied to each EEG segment to decompose them into the conventional EEG subbands, i.e., δ (1-4 Hz), θ (4-8 Hz), α (8-13 Hz), β (13-30 Hz) and γ (30- 40 Hz). In this case, S = 5 and the subband signals are related to the EEG subbands as follows:x _δ (n) = x ₀ (n), x _θ (n) = x ₁ (n), x _α (n) = x ₂ (n), x _β (n) = x ₃ (n) and x _γ (n) = x ₄ (n). [0060] The signal x ₅ (n) is not used since it does not have any kind of component as most of signal epochs related to this component have been filtered out. [0061] The Biorthogonal 3.5 is used as mother wavelet since it has proven to fit well with EEG signals from AD patients in [18-20]. The subband signals are upsampled to the original sampling frequency through the wavelet-based interpolation method [17]. [0062] The disclosed method according to the present disclosure, herein referred as Margolacs, of a discrete-time finite-energy signal x(n) is defined as the calculation of the magnitude of the inverse DWT of the logarithmic signal's DWT coefficients magnitude, i.e., where [0063] Figure 3 shows a schematic representation of an embodiment of EEG processing for obtaining calculated distances according to the present disclosure, wherein a selected wavelet multi-filtering process is applied to an EEG signal input, then, every resulting signal from the filtering multiband step suffered a decimation process followed by the module and the logarithm application, to attain the logarithmic magnitude of each EEG band signal. The Inverse Wavelet multi-filtering process followed by an interpolation task was applied to each EEG band signal logarithmic magnitude. The last step corresponds to the application of the module to obtain the magnitude of each resulting signal. [0064] Due to the logarithm, the disclosed calculation, herewith Margolacs, gets similar proprieties to the cepstrum but with a different transform in its conception that allows sliding analysis of non-stationary signals [17]. [0065] In the multiband analysis, the proposed disclosed method is calculated from each resulting EEG subband signal {δ, θ, α, β, γ}, by using S = 5, resulting in x ⁱ (n), where i = {δ, θ, α, β, γ} [0066] Several measures can be obtained from the disclosed method. In the present work, the distances are calculated for 5s-long segments of the 19 EEG channels of each subject. They are described below considering i = {δ, θ, α, β, γ} and j = {δ, θ, α, β, γ} as the subband indexes (i ≠ j) and N as the disclosed method signal length [0067] The normalized root mean square (RMS) distance is defined as [8]: wherel = 4.3429 and p = 2 are normalization factors [8]. The weight of the component [x ⁱ (1) −x ^j (1)] ² , herein named first Margolacs component, is reduced because, in general, it does not contain any important information about the EEG signal [8]. [0068] The normalized RMS distance without the first Margolacs component is defined as [19]: [0069] This distance also removes the first Margolacs component because, as mentioned, it usually does not contain any important information about the EEG signal. [0070] The Euclidean distance is defined as [22]: [0071] As the influence of the first Margolacs component is small, is practically proportional to [0072] The quefrency weighted Margolacs distance is defined as [14]: [0073] The quefrency rooted weighted Margolacs distance is defined as [22]: [0074] The quefrency squared weighted Margolacs distance is defined as [22]: [0075] The weighting factors in D4 _i,j , D5 _i,j and D6 _i,j aim to improve the prediction accuracy of the distances and, consequently, to find more abrupt differences between study groups comparing with the Euclidean distance D3 _i,j [22]. These weights are optimized according to known methods. [0076] In an embodiment, the ANN is a feed-forward multilayer perceptron with a hyperbolic tangent sigmoid transfer function (tansig) as activation function, Levenberg-Marquardt as the training algorithm, mean squared as the error function, output layer with 4 nodes, input layer with 60 nodes and finally one hidden layer with 30 nodes. Wherein the 60 nodes of the input layers represent the 6 defined distances per binary comparison between the computed disclosed method of the 5 EEG convention bands, and the 4 nodes of the output layer represent three stages of AD, i.e., MCI, ADM, and ADA, and the healthy control group, i.e., HC. [0077] Two types of training classification have been applied the global channel classification and the scalp level classification. [0078] In an embodiment, the global channel classification for indicating of presence for Alzheimer disease comprises the leave-one-out cross-validation used in this work, allows that if one features vector, Margolacs distances, of a specific EEG channel of a specific subject is selected for testing in the classification procedure, then the other 18 channel feature vectors of the same patient are automatically unused in the learning procedure, thereby ensuring data independence. A total of 722 iterations are done in the leave-one-out cross-validation, where 704 features vectors, 1 for testing and 703 for training, are used in each iteration. This type of training classification is important for medical diagnosis decision support at is provides a probability of diagnosis. [0079] In an embodiment, the scalp level classification for prospection over the channel: Normal leave-one-out-cross-validation that implied the training of 19 ANN with Margolacs distances extracted in each electrode analysis. With this the Alzheimer Scalp-level activity prospection maps, image generated by using EEGLab [7] can be made as a new entry is presented to the algorithm. [0080] The present disclosure presents several advantages, namely real-time processing of the algorithm and no feature selection step is needed for training the Artificial Neural Network. [0081] The real-time processing occurs due to the fact that the multi-band analysis is done within the disclosed method which reduces the computation time, avoiding the application of DWT and its inverse twice in sequence, e.g., for lacsogram algorithm computation [21], and due to the logarithm position on the signal processing. [0082] In an embodiment, the interface acquires data from any kind of electroencephalographic machine, adapted to different kind of frequency rates acquisition systems, thus it works with any kind of electroencephalograph system. [0083] Figure 4 shows a schematic representation of an embodiment of an interface, wherein 1 represents an indicator for AD, 2 represents power spectral density metrics, 3 represents Alzheimer Scalp-level activity prospection, 4 represents a recording data and the corresponding EEG conventional sub-bands plot. [0084] In an embodiment, the interface works in real-time, and it provides doctors a set of electroencephalography activity information by running the disclosed method. [0085] The interface gives users the following information: indicator of presence for Alzheimer disease, using a pre-trained, feed-forward multilayer perceptron, ANN, as a new entry is provided, thus applying the global channel classification; indicator/prospection over each EEG channel for Alzheimer disease, using a pre- trained, feed-forward multilayer perceptron, ANN, as a new entry is provided, thus applying the scalp level classification Alzheimer for prospection over a channel; metrics extracted from the Power Spectral Density of the signal, e.g., the frequencies indexes of 50 % power reached, 95 % power reached and the average frequency; recorded EEG data and the corresponding EEG conventional sub-bands. [0086] The topographic maps of the present solution can provide more information in a near future, to find more AD activity patterns, if used for training new machine learning models to understand the scalp activity differences as the disease progress. [0087] The term "comprising" whenever used in this document is intended to indicate the presence of stated features, integers, steps, components, but not to preclude the presence or addition of one or more other features, integers, steps, components, or groups thereof. [0088] The disclosure should not be seen in any way restricted to the embodiments described and a person with ordinary skill in the art will foresee many possibilities to modifications thereof. The above-described embodiments are combinable. [0089] The following claims further set out particular embodiments of the disclosure. [0090] References [1] Saeedeh Afshari and Mahdi Jalili. “Directed Functional Networks in Alzheimer’s Disease: disruption of Global and Local Connectivity Measures”. In: IEEE Journal of Biomedical and Health Informatics 21.4 (July 2017), pp. 949–955. DOI: 10.1109/JBHI.2016.2578954. [2] Haleh Aghajani et al. “Diagnosis of Early Alzheimer’s Disease Based on EEG Source Localization and a Standardized Realistic Head Model”. In: IEEE Journal of Biomedical and Health Informatics 17.6 (Nov.2013), pp.1039–1045. DOI: 10.1109/JBHI.2013.2253326. [3] Kwaku Akrofi et al. “Classification of Alzheimer’s disease and mild cognitive impairment by pattern recognition of EEG power and coherence”. In: IEEE International Conference on Acoustics, Speech and Signal Processing. Dallas, USA, Mar.2010, pp.606– 609. DOI: 10.1109/ICASSP.2010. [4] C. Besthorn et al. “EEG coherence in Alzheimer’s disease”. In: Electroencephalography and Clinical Neurophysiology 90.3 (Mar. 1994), pp. 242–245. DOI: 10.1016/0013- 4694(94)90095-7.5495193. [5] Massimo Buscema et al. “The IFAST model, a novel parallel nonlinear EEG analysis technique, distinguishes mild cognitive impairment and Alzheimer’s disease patients with high degree of accuracy”. In: Artificial Intelligence in Medicine 40.2 (June 2007), pp.127– 141. DOI: 10.1016/j.artmed.2007.02.006. [6] Raymundo Cassani and Tiago Falk. “Alzheimer’s Disease Diagnosis and Severity Level Detection Based on Electroencephalography Modulation Spectral “Patch” Features”. In: IEEE Journal of Biomedical and Health Informatics 24.7 (July 2019), pp. 1–12. DOI: 10.1109/JBHI.2019.2953475. [7] Arnaud Delorme and Scott Makeig. “EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis”. In: Journal of Neuroscience Methods 134.1 (Mar. 2004), pp. 9–21. DOI: 10.1016/j.jneumeth.2003.10.009. [8] A. Gray and J. Markel. “Distance measures for speech processing”. In: IEEE Transactions on Acoustics, Speech, and Signal Processing 24.5 (Oct. 1976), pp. 380–391. DOI: 10.1109/TASSP.1976.1162849. [9] C. Huang et al. “Discrimination of Alzheimer’s disease and mild cognitive impairment by equivalent EEG sources: a cross-sectional and longitudinal study”. In: Clinical Neurophysiology 111.11 (Nov. 2000), pp. 1961–1967. DOI: 10.1016/s1388- 2457(00)00454-5. [10] Saleha Khatun, Bashir I. Morshed, and Gavin M. Bidelman. “A Single-Channel EEG- Based Approach to Detect Mild Cognitive Impairment via Speech-Evoked Brain Responses”. In: IEEE Transactions on Neural Systems and Rehabilitation Engineering 27.5 (May 2019), pp.1063–1070. DOI: 10.1109 /TNSRE.2019.2911970. [11] Verner Knott et al. “Quantitative electroencephalography in Alzheimer’s disease: comparison with a control group, population norms and mental status”. In: Journal of Psychiatry & Neuroscience 26.2 (Mar.2001), pp.106–116. [12] Henrique S. Malvar. Signal Processing with Lapped Transforms. Norwood, MA: Artech House, 1992. [13] Co Melissant et al. “A method for detection of Alzheimer’s disease using ICA- enhanced EEG measurements”. In: Artificial Intelligence in Medicine 33.3 (Mar.2005), pp. 209–222. DOI: 10.1016/j.artmed.2004.07.003. [14] K.K. Paliwal. “On the performance of the quefrency-weighted cepstral coefficients in vowel recognition”. In: Speech Communication 1.2 (Jan. 1982), pp. 151–154.DOI: 10.1016/0167-6393(82)90034-6. [15] A. A. Petrosian et al. “Recurrent neural network-based approach for early recognition of Alzheimer’s disease in EEG”. In: Clinical Neurophysiology 112.2 (Aug.2001), pp.1378– 1387. DOI: 10.1016 /s1388-2457(01)00579-x. [16] Simon-Shlomo Poil et al. “Integrative EEG biomarkers predict progression to Alzheimer’s disease at the MCI stage”. In: Frontiers in Aging Neuroscience 5 (Oct.2013), p.58. DOI: 10.3389/fnagi.2013.00058. [17] O. Rioul and M. Vetterli. “Wavelets and Signal Processing”. In: IEEE Signal Processing Magazine 8.4 (Oct.1992), pp.14–38. DOI: 10.1109/79.91217. [18] Pedro Miguel Rodrigues et al. “Early Detection of Electroencephalogram Temporal Events in Alzheimer’s Disease”. In: Design, Development, and Integration of Reliable Electronic Healthcare Platforms. Ed. by Anastasius Moumtzoglou. Hershey, PA: IGI Global, 2017, pp.112–131. DOI: 10.4018/978-1-5225-1724-5.ch007. [19] Pedro Miguel Rodrigues et al. “Electroencephalogram Hybrid Method for Alzheimer Early Detection”. In: Procedia Computer Science 138 (2018), pp. 209–214. DOI: 10.1016/j.procs.2018.10.030. [20] Pedro Miguel Rodrigues et al. “Electroencephalogram Signal Analysis in Alzheimer’s Disease Early Detection”. In: International Journal of Reliable and Quality E-Healthcare 7.1 (Jan.2018), pp.40–59. DOI: 10.4018/IJRQEH.2018010104. [21] Pedro M. Rodrigues et al. “Lacsogram: A New EEG Tool to Diagnose Alzheimer’s Disease”. In: IEEE Journal of Biomedical and Health Informatics 25.9 (Sept.2021), pp. 3384–3395. DOI: 10.1109/jbhi.2021.3069789. [22] Y. Tohkura. “A weighted cepstral distance measure for speech recognition”. In: IEEE Transactions on Acoustics, Speech, and Signal Processing 35.10 (Oct.1987), pp.1414– 1422. DOI: 10.1109/TASSP.1987.1165058. [23] Martin Vetterli and Jelena Kovačević. Wavelets and Subband Coding. Englewood Cliffs, NJ: Prentice Hall, 1995. [24] Franćois Vialatte et al. “Early detection of Alzheimer’s disease by blind source separation, time frequency representation, and bump modelling of EEG signals”. In: Artificial Neural Networks: Biological Inspirations - ICANN 2005. Ed. by W. Duch et al. Vol. 3696. Lecture Notes in Computer Science. Berlin, Heidelberg: Springer, 2005, pp.683– 692. DOI: 10.1007/11550822_106.