Amplification and Filtering Techniques in Biomedical Signal Processing

Biomedical signals (e.g., ECG, EMG, EEG) are typically small in amplitude, contaminated by strong interference, and often corrupted by both analog-front-end (AFE) limitations and electrode/tissue effects. As a result, biomedical signal processing commonly applies a pipeline of: keywordAFE conditioning → keywordamplification → keywordfiltering → sampling/quantization. In analog systems, amplification is designed to achieve high gain with low drift/noise and strong keywordCMRR (common-mode rejection ratio). Filtering then constrains bandwidth, suppresses known noise bands (e.g., mains hum), and prevents aliasing into the band of interest before ADC. Instrumentation amplifiers (or equivalent fully differential front ends) are standard for extracting small differential signals while rejecting common-mode interference. A practical AFE architecture often uses an instrumentation amplifier (INA) and one or more filters, sometimes followed by a programmable-gain amplifier (PGA).2

1) Amplification in Biomedical Signal Processing

1.1 Why amplification is non-trivial in biopotentials

Biopotential measurements are formed by a differential voltage between electrodes, often with a large common-mode component. Even when the desired signal is in the millivolt range, electrode potentials can include much larger DC offsets and common-mode interference; without careful design, amplifiers can saturate and/or distort the waveform. A common explanation in ECG instrumentation is that tissue-electrode DC potentials (near $0\ \text{Hz}$) can be orders of magnitude larger than the cardiac signal and must be removed/managed (often by high-pass filtering in addition to amplification).

In general, biomedical amplification must balance:

• keywordCMRR
• input impedance / electrode loading effects
• low input-referred noise and low drift
• bandwidth aligned to the diagnostic frequency content
• stability in fully differential or single-ended measurement configurations

The instrumentation amplifier is a high-performance extension of a differential amplifier, specifically designed to measure small differential voltages with high common-mode rejection, high input impedance, low drift, and low noise—making it well suited to ECG/EMG/EEG acquisition.

1.2 Instrumentation amplifier principles and CMRR

Instrumentation amplifiers suppress common-mode noise by amplifying the difference between electrode signals rather than the average. A key figure of merit is keywordCMRR, commonly expressed as the ratio of differential gain to common-mode gain.

Key idea: High CMRR reduces the impact of mains interference and electrode offset—both of which often appear as common-mode (or partially common-mode) components at the amplifier inputs.

Biomedical designs frequently implement fully differential analog front ends (rather than only single-ended) because they help improve both CMRR and power-supply rejection (PSRR), and they can simplify differential ADC interfacing. For example, EMG AFE designs often use an INA plus a programmable gain stage along with filtering, implemented in a fully differential topology.

2) Filtering Techniques (Analog and Digital)

2.1 Filtering targets in biomedical signals

Filtering is used to suppress:

• keywordbaseline wander (very low frequencies)
• keywordpower line interference
• keywordEMG/muscle artifacts (often higher-frequency content, depending on modality)
• keywordelectrode/motion artifacts
• out-of-band energy that can alias into the band of interest

For ECG specifically, baseline wander is often addressed with a high-pass filter around $0.5\,\text{Hz}$, while mains hum is reduced with a 50/60 Hz notch filter, and high-frequency noise is reduced with a low-pass filter. However, as noted above, filter choices can alter waveform morphology and impact clinical interpretation.2

2.2 Anti-alias filtering before ADC (why it matters)

When you sample a continuous-time signal, frequencies above the Nyquist frequency can fold (“alias”) into the lower frequencies, corrupting measurement. An anti-aliasing filter (AAF) restricts bandwidth before the sampler to better satisfy the sampling theorem’s conditions. A practical AAF trades off ideal sharp cutoffs vs implementable attenuation and phase/distortion limits.

A key practical consequence: wide ADC bandwidth allows more noise from multiple Nyquist zones to alias into the output, raising the noise floor; therefore, anti-alias filtering is often used to reduce out-of-band noise before digitization.

2.3 Common filter classes used in biomedical systems

Biomedical pipelines frequently combine multiple filter types:

• High-pass filters: remove DC offset and baseline wander (e.g., around $0.5\,\text{Hz}$ in ECG contexts).2
• Low-pass filters: suppress high-frequency noise (e.g., muscle artifacts), with cutoffs tailored to modality.
• Band-pass filters: keep the diagnostic band while suppressing both low- and high-frequency interference; ECG morphology is concentrated roughly in low tens of Hz bands.
• Notch filters: suppress narrowband interference at $50/60\,\text{Hz}$ (and harmonics), often implemented digitally when the interference frequency is known and stable.
• FIR vs IIR
  • FIR filters are often selected when linear phase or guaranteed stability matters and more computation is acceptable.
  • IIR filters can achieve sharp magnitude response with lower order but may have nonlinear phase and stability considerations.

While biomedical-specific FIR/IIR comparisons vary by implementation, general DSP practice notes that FIR filters are feed-forward and always stable, while IIR can be more efficient but has different tradeoffs. Biomedical ECG denoising literature commonly implements both FIR and IIR designs to remove baseline and power-line noise.

3) Putting It Together: Representative AFE + Filtering Pipeline

3.1 Typical biomedical analog front-end structure

A common AFE block diagram is: electrode → low-noise instrumentation amplifier → gain stage → analog band-limiting filters → ADC. For EMG, a preprocessing block is often described as INA + programmable gain amplifier (PGA) plus a filter, forming the analog front-end that conditions the weak sensor output for further processing.

3.2 Why fully differential design is prevalent

Fully differential approaches can improve rejection metrics and facilitate downstream differential digitization. EMG AFE examples describe an INA and PGA implemented in fully differential topology to achieve high common-mode rejection and PSRR behavior in low-voltage designs.

3.3 Filter parameter selection depends on clinical feature preservation

ECG filtering is especially sensitive for ST-segment and J-point morphology. Evidence shows that high-pass filters around $0.5\,\text{Hz}$ can introduce clinically significant changes in ST-segment interpretation, even when low-pass cutoffs differ. Therefore, biomedical filtering is best treated as a constraint-optimization problem: suppress artifacts while preserving diagnostic features.