UNIT 4 NOTES (BIOMEDICAL SIGNAL PROCSSING)

🧠 1. EEG: Neurological Signal Processing

Electroencephalography (EEG) is a technique used to record the electrical activity of the

brain using electrodes placed on the scalp. The brain consists of millions of neurons that
generate electrical signals, which are captured by these electrodes. The recorded EEG signal
is very weak in amplitude and contains noise from various sources like eye movements and
muscle activity. Therefore, signal processing techniques such as amplification, filtering, and
analysis are applied to extract useful information. EEG is widely used in medical diagnosis,
especially for studying brain disorders, sleep patterns, and epilepsy.

Brain Activity
↓
Electrodes on Scalp
↓
Amplifier
↓
Filter (remove noise)
↓
Signal Processing
↓
Display / Analysis
Key Points:

 Signal is very small (µV)

 Changes with time (non-stationary)

 Contains noise

🧠 2. Characteristics of EEG Signal

EEG signals have several important characteristics that make them complex to analyze. They
are non-stationary in nature, meaning their statistical properties change with time. The
amplitude of EEG signals is very low, typically in the range of microvolts, so amplification is
required. These signals are often contaminated with noise from eye movements, muscle
activity, and external interference. EEG signals also appear random and unpredictable.
Additionally, they are classified based on frequency ranges, which correspond to different
mental states such as sleep, relaxation, and alertness.
EEG Signal
↓
Low Amplitude
↓
Noise Present
↓
Time Varying
↓
Different Frequency Bands

Easy Points:

 Non-stationary signal

 Low amplitude

 Noisy signal

 Random in nature

 Frequency varies with brain state

🧠 3. Brain Waves Pattern

EEG signals are divided into different frequency bands known as brain waves, which
represent various states of brain activity. Delta waves (0.5–4 Hz) are associated with deep
sleep and have high amplitude. Theta waves (4–8 Hz) occur during light sleep or drowsiness.
Alpha waves (8–13 Hz) are observed when a person is relaxed and calm, usually with eyes
closed. Beta waves (13–30 Hz) are present during active thinking and concentration. As the
frequency increases from delta to beta, the amplitude generally decreases. These patterns
help in understanding the mental and physiological condition of a person.

Wave Frequency State

Delta 0.5–4 Hz Deep sleep

Theta 4–8 Hz Light sleep

Alpha 8–13 Hz Relaxed

Beta 13–30 Hz Active

Flow chart
Brain Activity
↓
Signal Frequency
↓
Delta → Theta → Alpha → Beta
↓
Different Mental States

🧠 4. Linear Prediction Theory

Linear Prediction Theory is a method used to estimate the current value of a signal based on
its past values. In EEG signal processing, it assumes that the present sample can be
expressed as a linear combination of previous samples plus an error term. The main
objective is to minimize this prediction error. This theory is useful for modeling EEG signals
and is widely used in autoregressive (AR) spectral estimation methods. It helps in reducing
redundancy and efficiently representing the signal.

Formula:

𝑥(𝑛) = −𝑎1 𝑥(𝑛 − 1) − 𝑎2 𝑥(𝑛 − 2) + 𝑒(𝑛)

Flow chart

Past EEG Samples

↓
Multiply with Coefficients
↓
Add Values
↓
Predicted Present Value
↓
Error Calculation

🧠 5. Sleep EEG

Sleep EEG refers to the study of brain activity during different stages of sleep. Sleep is
divided into several stages including awake, N1, N2, N3, and REM. In the awake state, beta
waves are dominant. As the person enters sleep, theta waves appear in stage N1. Stage N2 is
characterized by sleep spindles and K-complexes. In deep sleep (stage N3), delta waves
dominate, indicating very slow brain activity. REM sleep shows brain activity similar to the
awake state and is associated with dreaming. EEG helps in identifying these stages and
analyzing sleep disorders.

Sleep Stages:
  Awake → Beta

 N1 → Theta

 N2 → Spindles

 N3 → Delta

 REM → Active

Flowchart:

Awake (Beta)
↓
Stage N1 (Theta)
↓
Stage N2 (Spindles)
↓
Stage N3 (Delta)
↓
REM Sleep

🧠 6. Dynamics of Sleep/Wake Transition

The transition between sleep and wakefulness is a gradual process reflected in EEG signals.
When a person moves from wakefulness to sleep, the brain activity slows down, and alpha
waves are gradually replaced by theta waves. As sleep deepens, delta waves appear. During
waking up, the reverse process occurs, where delta and theta waves are replaced by alpha
and beta waves. This transition is important for understanding sleep patterns and diagnosing
sleep-related problems.

Flowchart:

Awake
↓
Alpha Waves
↓
Theta Waves
↓
Sleep

(Wake-up is reverse)
🧠 7. Study of Brain Wave Patterns

The study of brain wave patterns involves analyzing EEG signals to understand different
mental states and neurological conditions. These patterns vary depending on activities such
as sleeping, thinking, or being relaxed. Analysis can be done in the time domain, frequency
domain, or both using time-frequency methods. By studying these patterns, doctors and
researchers can detect abnormalities, monitor brain function, and diagnose disorders like
epilepsy and sleep disturbances.

Purpose:

 Identify brain state

 Detect disorders

Methods:

 Time analysis

 Frequency analysis

Flowchart:

EEG Signal
↓
Analysis Method
↓
Pattern Detection
↓
Brain State Identification

Brain waves-

 Delta → deep sleep (slowest)

 Theta → light sleep

 Alpha → relaxed

 Beta → active thinking

Important Line:

👉 “Frequency increases from Delta to Beta, while amplitude decreases”

🧠 8. Epilepsy: Transition, Detection and Estimation

Epilepsy is a neurological disorder characterized by sudden and abnormal electrical activity

in the brain. In EEG recordings, this appears as spikes, sharp waves, or spike-and-wave
patterns. The transition from normal brain activity to seizure activity can be identified by
sudden changes in the signal. Detection methods involve analyzing EEG signals to identify
these abnormal patterns. Estimation includes determining the duration, intensity, and
frequency of seizures. EEG plays a crucial role in diagnosing and monitoring epilepsy.

Flow chart

Normal EEG
↓
Sudden Change
↓
Abnormal Spikes
↓
Seizure Detection
↓
Estimate Duration & Severity

Easy Explanation:

Epilepsy is a chronic neurological disorder characterized by recurrent, unprovoked seizures

caused by sudden, excessive electrical discharges in brain cells.

EEG Shows:

 Spikes

 Sharp waves

📊 9. EEG Analysis by Spectral Estimation

Spectral estimation is used to analyze EEG signals in the frequency domain by determining
how the signal power is distributed across different frequencies. Since EEG signals are
complex in the time domain, transforming them into the frequency domain makes analysis
easier. Various methods like periodogram, Blackman-Tukey, AR, and others are used for
spectral estimation. This analysis helps in identifying dominant brain waves and detecting
abnormalities.
Easy Idea:

Convert signal → frequency domain to see power distribution

Flowchart:

EEG Signal
↓
Apply Method
↓
Frequency Spectrum
↓
Power vs Frequency

📊 10. Periodogram Method

The periodogram method is a basic technique used to estimate the power spectral density
of a signal. It is calculated by taking the Fourier transform of the signal and then computing
the squared magnitude. Although it is simple and easy to implement, the periodogram
method suffers from high variance and produces noisy results. Therefore, it is not very
reliable for accurate spectral estimation but is useful for basic analysis.

Easy Explanation:

A periodogram in biomedical signal processing is a foundational non-parametric method

used to estimate the power spectral density (PSD) of a signal, identifying the dominant
frequencies and their intensities. It transforms time-domain data (e.g., ECG, EEG) into the
frequency domain using the Fast Fourier Transform (FFT), highlighting periodic components
and cyclical behaviors, such as brain wave oscillations

Flowchart:

EEG Signal
↓
FFT
↓
Magnitude Square
↓
Power Spectrum
📊 11. Blackman-Tukey Method

The Blackman-Tukey method improves spectral estimation by using the autocorrelation

function of the signal. First, the autocorrelation of the EEG signal is calculated, and then a
window function is applied to reduce errors. Finally, the Fourier transform is taken to obtain
the power spectrum. This method provides smoother results compared to the periodogram
but has lower frequency resolution.

📊 12. Autoregressive (AR) Method

The autoregressive (AR) method models the EEG signal as a linear combination of its past
values and an error term. It provides high-resolution spectral estimation, even with short
data lengths. The AR model is widely used in EEG analysis because it can accurately
represent the signal and detect important frequency components. However, selecting the
correct model order is important for obtaining accurate results.

Easy Explanation:

Uses past values to model signal

Flowchart:

Past Values
↓
Apply Coefficients
↓
Prediction
↓
Error Minimization
↓
Power Spectrum

📊 13. Moving Average (MA) Method

The moving average (MA) method represents the signal as a linear combination of present
and past noise values. It is mainly used for modeling noise in signals. Although it is simpler
than other methods, it is less flexible and not commonly used alone for EEG analysis.
However, it forms an important part of combined models like ARMA.

Easy Explanation:

Uses past noise values

Flowchart:

Noise Signal
↓
Weighted Sum
↓
Output Signal
↓
Spectrum

📊 14. ARMA Method

The ARMA (Autoregressive Moving Average) method is a combination of AR and MA models.

It considers both past signal values and past noise values to model the EEG signal. This
makes it more flexible and accurate compared to individual AR or MA models. However, the
complexity of the ARMA method is higher, and parameter estimation is more difficult.

Easy Explanation:

Combination of AR + MA

Flowchart:

Past Values + Noise

↓
AR + MA Model
↓
Signal Output
↓
Spectrum

📊 15. Maximum Entropy Method (MEM)

The Maximum Entropy Method is a spectral estimation technique based on the AR model. It
works by maximizing the entropy, which means it assumes the least prior information about
the signal. This results in a power spectrum with sharp peaks and high resolution. MEM is
especially useful in EEG analysis for detecting specific frequency components, but it may
sometimes produce misleading results if not used carefully.

Easy Explanation:
Gives best resolution using AR model

Flowchart:

EEG Signal
↓
AR Modeling
↓
Maximize Entropy
↓
Sharp Spectrum

📊 16. Maximum Likelihood Method

The Maximum Likelihood Method estimates the parameters of a signal model by maximizing
the likelihood function, which represents the probability of observing the given data. This
method provides very accurate results and is based on statistical principles. However, it
involves complex calculations and requires more computational effort, making it less
commonly used in simple EEG analysis.

Easy Explanation:

Uses probability to estimate parameters

Flowchart:

EEG Data
↓
Assume Model
↓
Likelihood Calculation
↓
Maximize Likelihood
↓
Best Parameters
COMPARISON BETWEEN MEM AND MLM

Feature MEM MLM

1. Full Form Maximum Entropy Method Maximum Likelihood Method

2. Basis Entropy (randomness) Probability (likelihood)

3. Model Type AR model based Statistical model

4. Resolution Very high (sharp peaks) High accuracy

5. Complexity Moderate Very high

6. Computation Faster Slower

7. Accuracy Good Very accurate

8. Risk Overfitting possible Less overfitting

9. Application EEG spectral analysis Parameter estimation

COMPARISON BETWEEN AR, MA & ARMA METHOD

Feature AR MA ARMA

Autoregressive Moving
Full Form Autoregressive Moving Average
Average

Depends On Past values Past noise Both

Model
Both combined
Equation

Complexity Low Low High

Flexibility Moderate Low High

Accuracy Good Less Very good

Feature AR MA ARMA

Only MA
Parameters Only AR coefficients Both
coefficients

Usage in EEG Very common Rare Advanced analysis