(SEM VIII ) THEORY EXAMINATION 2022-23 BIOMEDICAL SIGNAL PROCESSING

SECTION A

(Attempt all | 2 × 10 = 20 Marks)

(a) General purpose microprocessor vs DSP

A general-purpose microprocessor is designed for a wide range of applications and performs sequential operations, whereas a DSP (Digital Signal Processor) is optimized for real-time signal processing with features like fast multiply-accumulate operations and parallel architecture.

(b) Dominant frequencies in sleep EEG

Delta (0.5–4 Hz): Deep sleep                                         Theta (4–7 Hz): Light sleep

Alpha (8–13 Hz): Relaxed wakefulness                          Beta (13–30 Hz): Alert state

(c) QRS complex

The QRS complex represents ventricular depolarization in an ECG signal and indicates the contraction of the ventricles.

(d) Use of FAN algorithm

The FAN (Floating Amplitude Normalization) algorithm is used to detect QRS complexes in ECG signals by enhancing peak detection under varying signal amplitudes.

(e) EP estimation

EP (Evoked Potential) estimation refers to extracting small stimulus-related signals from EEG by averaging repeated responses.

(f) Different patterns of brain waves

Delta, Theta, Alpha, Beta, and Gamma waves.

(g) Action potential and resting potential

Resting potential is the voltage difference across a cell membrane when inactive. Action potential is the rapid change in membrane potential during nerve impulse transmission.

(h) Need for data reduction

Data reduction reduces storage requirements, transmission bandwidth, and computational complexity while preserving essential diagnostic information.

(i) Sleep EEG

Sleep EEG records electrical brain activity during sleep to analyze sleep stages and disorders.

(j) Types of biomedical signals

ECG, EEG, EMG, EOG, blood pressure signals, respiratory signals, and evoked potentials.

SECTION B

(Attempt any THREE | 10 × 3 = 30 Marks)

2(a) Objectives of biomedical signal analysis

Biomedical signal analysis aims to extract useful physiological information, detect abnormalities, assist diagnosis, monitor patient health, and support treatment planning.
Objectives include noise removal, feature extraction, pattern recognition, compression, and interpretation of physiological signals.
(Block diagram explanation: signal acquisition → preprocessing → feature extraction → analysis → diagnosis)

2(b) Portable arrhythmia monitor

A portable arrhythmia monitor continuously records ECG signals over long durations. It detects irregular heart rhythms using electrodes, amplifiers, signal processors, memory units, and display systems. These devices help diagnose intermittent cardiac abnormalities.

2(c) Three approaches for QRS detection

Amplitude thresholding: Detects peaks exceeding a threshold

Slope-based detection: Uses rapid slope changes in QRS

Digital filtering: Enhances QRS frequency components

2(d) EEG analysis using spectral estimation

EEG analysis uses methods like:                                 FFT-based spectral analysis

Autoregressive (AR) modeling                               Time-frequency analysis
These techniques estimate power distribution across frequency bands to study brain activity.

2(e) Adaptive wavelet detection

Adaptive wavelet detection uses wavelet transforms to identify transient features like QRS complexes. Overlapping wavelets are detected by matching signal patterns at multiple scales, improving accuracy in noisy signals.

SECTION C

3(a) Run Length Encoding (RLE)

Run Length Encoding reduces data by representing consecutive repeated values as a single value and count.
Example: 111122 → (1,4)(2,2).
It is simple and effective for slowly varying biomedical signals.

3(b) Discrete signal epochs

Biomedical signals are divided into time segments (epochs) and correlated with physiological events such as heartbeats or brain responses to analyze functional relationships.

4(a) Importance of signal averaging

Signal averaging enhances weak signals by reducing uncorrelated noise.
In ECG averaging, repeated beats are aligned and averaged to improve signal-to-noise ratio.
(Block diagram + flow chart explanation: acquisition → alignment → averaging → output)

4(b) Lossless and lossy data compression

Lossless: No information loss (RLE, Huffman coding)

Lossy: Some information loss (AZTEC, wavelet compression)

Classification: Direct data reduction, transform-based, predictive coding.
Lossless algorithm example: Run Length Encoding.

5(a) AZTEC reconstructed waveform issues

AZTEC reconstruction produces staircase-like waveforms, distorting ECG morphology.
This is unacceptable to cardiologists as it alters diagnostic features.
Improvements include slope interpolation and adaptive thresholding.

5(b) Markov model for sleep EEG

Markov models represent sleep stages as probabilistic state transitions.
They work well for stage sequencing but fail when external factors (stress, drugs) influence sleep patterns.

6(a) AZTEC encoding problem

(i) Data reduction:
Original data: continuous samples                Encoded data: pairs → significant reduction (typically >80%)

(ii) Peak-to-peak amplitude:
Maximum = +50, Minimum = −6                 Peak-to-peak = 56 units

6(b) Removal of baseline wander and power-line interference

Baseline wander: Removed using high-pass filtering or polynomial fitting

Power-line interference: Removed using notch filters at 50/60 Hz

7(a) Detection and estimation of Epilepsy

Epilepsy detection involves identifying abnormal EEG spikes, rhythmic discharges, and seizure patterns. Techniques include time-domain analysis, frequency analysis, wavelet transforms, and pattern recognition.

7(b) Signal averaging SNR problem

Noise amplitude = 4 × signal                     Required SNR = 4:1

SNR improves as √N                                   N=16⇒N=256\sqrt{N} = 16 \Rightarrow N = 256N​=16⇒N=256

Answer: 256 sweeps are required.