What Is Biomedical Signal Processing?

Biomedical signal processing is a discipline of biomedical engineering and electrical engineering concerned with the extraction of clinically and physiologically meaningful information from signals produced by the body. These signals include the electrical potentials of the heart (ECG), brain (EEG), and muscles (EMG); acoustic signals such as heart sounds and lung breath sounds; optical signals from pulse oximetry and photoplethysmography; and biochemical concentration time series from continuous glucose monitors. The field addresses the full pipeline from raw signal acquisition through noise suppression, feature extraction, classification, and interpretation. Biomedical signal processing draws on linear systems theory, statistical estimation, information theory, and increasingly on machine learning, with its outputs informing clinical diagnosis, patient monitoring, and the design of closed-loop therapeutic systems.

Time-Frequency Analysis and Feature Extraction

Many biomedical signals are non-stationary: their frequency content changes over time in ways that carry diagnostic information. The short-time Fourier transform partitions a signal into overlapping windows and computes the spectrum of each, producing a spectrogram that reveals transient spectral events. Wavelet transforms provide a multiresolution alternative: the continuous wavelet transform (CWT) decomposes a signal into a two-dimensional time-scale representation that resolves high-frequency transients at fine time resolution while tracking low-frequency trends at coarser resolution. QRS complexes in the ECG and spike-and-wave discharges in the EEG are characteristic events that time-frequency methods locate and characterize. An IEEE conference paper on time-frequency signal processing techniques for biomedical classification compares wavelet, Hilbert-Huang, and empirical mode decomposition methods across clinical datasets. A PMC survey of EEG signal processing for biomedical applications catalogs feature sets drawn from the time, frequency, and wavelet domains that feed into seizure detection, sleep staging, and brain-computer interface classifiers.

Neurophysiological Signal Processing

Neurophysiological signal processing covers EEG, electrocorticography (ECoG), and extracellular neural recordings, each carrying information at different spatial and temporal resolutions. EEG analysis identifies rhythmic activity in defined frequency bands: delta (1 to 4 Hz), theta (4 to 8 Hz), alpha (8 to 13 Hz), beta (13 to 30 Hz), and gamma (above 30 Hz), each associated with distinct cognitive states and pathological signatures. Spike sorting algorithms applied to microelectrode array recordings isolate the action potentials of individual neurons by clustering waveform shapes in feature space, enabling the study of neural population codes. Brain-computer interfaces decode motor intentions from primary cortex ECoG signals at tens of milliseconds latency to control prosthetic limbs or communication devices. An IEEE journal review of EEG signal processing for medical diagnosis and monitoring covers artifact rejection, source separation using independent component analysis, and classification pipelines applied to disorders of consciousness, epilepsy, and attention deficit conditions.

Adaptive Filtering and Artifact Removal

Biomedical signals are routinely contaminated by motion artifacts, power-line interference at 50 or 60 Hz, electromyographic noise, and electrode contact transients. Adaptive filters using the least mean squares or recursive least squares algorithm track and suppress noise with statistical properties that change over time, without requiring a fixed noise model. Independent component analysis (ICA) separates mixed bioelectrical sources by statistical independence, recovering clean cardiac or ocular signals from EEG channels that contain multiple superimposed contributions. Fall detection, a practical application of biomedical signal processing, combines accelerometer and gyroscope signals with threshold-based or machine-learning classifiers to distinguish fall events from normal activities, reducing mortality risk in elderly populations monitored by wearable devices.

Applications

Biomedical signal processing has applications in a wide range of disciplines, including:

• Clinical electrocardiography and cardiac arrhythmia detection
• Epilepsy monitoring and seizure prediction
• Brain-computer interfaces for motor and communication rehabilitation
• Sleep staging and polysomnographic analysis
• Wearable fall detection in elderly care
• Fetal heart rate monitoring during labor