From Steps to Emotions: A Taxonomy of Wearable Sensor Modalities

Wearable sensing spans a progression from movement and body mechanics to physiology, psychophysiology, and chemistry. This taxonomy maps what each modality detects, how mature and costly it appears in the supplied evidence, where it is useful, and why combining sensors is often harder than adding another sensor.

The evidence supports qualitative comparisons rather than universal performance rankings: the supplied reviews provide few modality-specific accuracy figures or prices. Accordingly, “cost” below means an evidence-supported design tier, while “accuracy” emphasizes known error sources and validation needs.

1. Taxonomy at a Glance

2. From Movement to Physiological State

IMUs provide a relatively direct route from body movement to measurable kinematics. Their principal limitation is not simply sensor noise: orientation estimates can drift or saturate, which is why accelerometer and gyroscope fusion is used to improve longer-duration estimates.^[22][23] The supplied evidence names motion analysis, consumer fitness, clinical monitoring, industrial safety, and human-motion analysis as application areas for wearable sensing more broadly.^[24]

Optical and electrical modalities add physiological information. PPG measures heart rate and oxygen saturation, and combining PPG with ECG supports pulse-transit-time-based blood-pressure estimation.^[25][26] ECG, EMG, and EEG use skin-contact electrodes, while bioimpedance is another measurable electrical quantity.^[27][28][29] Thermal sensing contributes skin temperature, and pressure arrays contribute spatially resolved pressure information.^[30][31]

For emotion-related inference, the evidence supports a multimodal framing rather than a one-sensor interpretation. EDA appears as one channel in multimodal psychophysiological monitoring, and multimodal systems may combine ECG, PPG, accelerometry, EDA, temperature, EEG, and EMG.^[32] The important distinction is that these channels provide correlated physiological context; they do not, by themselves, establish a unique emotional state.

3. Biochemical and Molecular Detectors: Current to Future

Wearable biochemical systems sample accessible biofluids or minimally invasive compartments. Soft microfluidic platforms use electrochemical, colorimetric, optical, and related methods to detect metabolites, hormones, electrolytes, hydration status, and other biomarkers in sweat and other biofluids.^[33][34] The broader biomarker set reported in the supplied review includes glucose, lactate, cortisol, pH, uric acid, inflammation markers, alcohol, nutrients, and vitamins, with sampling discussed for sweat, saliva, tears, breath, and interstitial fluid.^[35][36]

Sweat sensing is attractive because sweat is noninvasively accessible and can be sampled continuously, but low production and contamination from skin or the environment limit reliability. Some sweat concentrations may correlate with blood levels, yet low-abundance hormones, proteins, and peptides are often difficult to detect and poorly correlated with blood measurements.^[37][38][39] These limitations make sweat useful for exercise monitoring, hydration and metabolic monitoring, decentralized health monitoring, and exploratory diagnostics, but they do not make it a universal substitute for blood testing.^[40][41]

Interstitial-fluid systems use small probes or microneedles beneath the skin and are described as minimally invasive; blood remains the clinical gold standard in the cited review.^[42][43] Microneedle patches are being developed for glucose, electrolytes, drug levels, lactate, and other biomarkers, including transdermal sensing and possible closed-loop therapeutic applications.^[44][45]

Readiness is heterogeneous. The evidence describes some sweat patches and FreeStyle Libre as currently available; graphene sweat sensors as being in prototyping and clinical testing; biodegradable skin sensors as early laboratory prototypes; microneedle patches as in preclinical trials and regulatory review; paper-based sensors as in laboratory testing or preparation for pilot studies; and smart tattoos as still in developmental testing.^{[46][47][48][49]} No supplied source provides a complete numerical comparison of sensitivity, specificity, accuracy, or monetary cost across these technologies.^[50]

4. Accuracy Is a System Property

Wearable accuracy changes with the body, placement, movement, contact, environment, and time. Soft-tissue motion can change the relationship between a sensor and the underlying anatomy, while skin stretches, compresses, and shears across different body regions. These effects complicate calibration, electrode placement, and transfer of algorithms between users and locations.^[51][52][53]

Motion artifacts can exceed the physiological signal substantially: the cited review reports that they are typically at least ten times greater than biosignals in EEG and PPG systems.^[54] In PPG, sensor displacement is a major distortion source; EEG can also be affected by muscle and eye signals, electromagnetic interference, and movement of the measurement system.^[55] Because motion and physiological spectra can overlap, filtering alone may be insufficient. Mechanical fixation, improved electrodes and analog circuits, inertial references, adaptive filtering, source separation, and learning-based compensation are among the mitigation approaches identified in the evidence.^[56]

Biochemical accuracy has a different failure profile. Electrochemical sensors can lose sensitivity or selectivity through irreversible binding, degradation, biofouling, temperature changes, mechanical strain, prolonged fluid exposure, microcracks, and degradation of redox mediators. Optical and colorimetric systems can be affected by lighting, motion, pH, biofluid composition, and inconsistent fluid handling.^[57][58]

5. Integration Challenges and Design Priorities

Conclusion: A Practical Map of the Field

The taxonomy moves from relatively direct physical signals, such as acceleration and temperature, toward increasingly context-dependent physiological, psychophysiological, and biochemical signals. IMUs and established electrical or optical channels support mature motion and physiological monitoring, while biochemical and molecular systems add potentially richer chemical information but face greater sampling, stability, manufacturing, and validation barriers.^{[71][72][73][74]}

The most defensible design principle from the supplied evidence is not “more sensors,” but quality-aware integration: align channels in time and space, monitor signal quality, detect artifacts, manage energy and comfort, and validate the resulting interpretation against meaningful outcomes. Multimodal fusion can improve resilience, yet it can also reduce performance when weak or incompatible signals are combined without fault handling.^[75][76][77]