The Overtone Frontier: Chip-Scale Sensors, Chemometrics, and the Long Chase for Noninvasive NIR Diagnostics

Near-infrared (NIR) spectroscopy, built on the weak overtone and combination-band absorptions first exploited commercially in the 1970s, is now being reshaped by chip-scale MEMS sensors, rapid field calibration methods, and increasingly capable chemometric and machine-learning models.

Summary

Handheld and consumer-facing NIR devices are proliferating in agriculture, food quality, and pharmaceutical screening even as the long-sought goal of fully noninvasive glucose monitoring remains elusive.

This article reviews recent NIR instrumentation, patents, and research, featuring perspectives from award-winning researchers and the developers building the next generation of miniaturized NIR sensors.

Introduction

Near-infrared spectroscopy occupies a curious niche in analytical chemistry: it measures the weak overtone and combination bands of fundamental mid-infrared vibrations, absorptions that are typically ten to a thousand times less intense than their mid-infrared counterparts. For decades that weakness was considered a liability. In practice, it has proven to be NIR's greatest asset, allowing measurements through packaging, in thick or scattering samples, and with minimal sample preparation, conditions under which mid-infrared and Raman techniques often struggle.

Karl Norris's pioneering work applying NIR to agricultural moisture and protein analysis in the 1960s and 1970s established the technique's first major commercial foothold, and chemometric regression methods such as partial least squares later became inseparable from practical NIR analysis, since NIR bands are so broad and overlapping that direct peak assignment is rarely possible. Professor Yukihiro Ozaki, professor emeritus at Kwansei Gakuin University and recipient of the 2025 Ellis R. Lippincott Award from Optica, the Society for Applied Spectroscopy, and the Coblentz Society, has spent much of his career elucidating exactly what those overlapping bands mean.¹

“My most impactful contributions to the NIR field were fundamental studies of NIR spectroscopy, studies of overtones, combination modes, and anharmonicity, [and] those of spectral analysis methods such as chemometrics, two-dimensional correlation spectroscopy, and anharmonic quantum chemical calculations.”²

— Professor Yukihiro Ozaki, Kwansei Gakuin University, 2025 Ellis R. Lippincott Award recipient

This article surveys how chip-scale sensor manufacturing, rapid field calibration methods, and advanced chemometrics are extending NIR spectroscopy's reach, from grain silos to the long-elusive goal of noninvasive medical diagnostics.

From FT-NIR to Chip-Scale MEMS Sensors

Ozaki himself recalled one of the field's pivotal instrumental transitions, the shift from filter- and grating-based dispersive NIR instruments to Fourier-transform NIR spectrometers.

“We bought an FT-Raman instrument in 1992 and found that this could be used as an FT-NIR spectrometer. I think we may be one of the first to start using FT-NIR spectroscopy.”²

— Professor Yukihiro Ozaki, Kwansei Gakuin University

Three decades later, NIR instrumentation has undergone an equally significant transition: miniaturization onto semiconductor-scale chips. Si-Ware Systems' NeoSpectra platform was among the first fully integrated, chip-sized NIR spectral sensors targeted at industrial and consumer markets, using a MEMS-based Michelson interferometer fabricated with standard semiconductor processes to shrink an FT-NIR spectrometer from a benchtop instrument into a component small enough to embed in handheld and portable devices.³ Competing chip-scale approaches, including ams OSRAM's AS7421 hyperspectral NIR sensor family, use linear variable filter arrays over photodiode banks to achieve similar miniaturization with a different optical architecture, reflecting active competition among multiple sensor technologies for the same emerging embedded-spectroscopy market.⁴

Faster Field Calibration for Fabry-Perot Systems

Miniaturized NIR sensors are only useful if they can be calibrated quickly and reliably outside a controlled laboratory environment, historically one of the technique's persistent practical bottlenecks. A U.S. patent granted to Hinalea Imaging Corp. (formerly Trutag Technologies) in January 2024 addresses this directly. Invented by Guocai Shu and Alexandre Fong, the patent describes a field calibration method for tunable Fabry-Perot etalon-based spectral measurement systems that uses a limited set of field material targets combined with previously stored full laboratory calibrations to generate a rapid reconstruction matrix.⁵ According to the patent's technical disclosure, this approach can reduce calibration time to roughly one-thousandth, one-hundredth, or one-tenth of conventional full recalibration procedures, while enabling near-real-time hyperspectral measurement at approximately ten frames per second, compared to five to ten seconds per frame with earlier systems.⁵ Faster field calibration directly addresses one of the most persistent practical bottlenecks limiting deployment of hyperspectral and NIR imaging systems outside controlled laboratory settings.

Agriculture and Food Quality: NIR's Proving Ground

Agriculture remains NIR spectroscopy's most mature and highest-volume application, and it continues to benefit from continued instrument miniaturization. Handheld MicroNIR instruments have demonstrated reliable determination of protein content in sorghum grain samples under field conditions, correlating closely with reference wet-chemistry methods while eliminating the delay of laboratory submission.⁶ Purpose-built portable analyzers, such as the GrainSense handheld grain quality analyzer, now allow farmers and grain buyers to assess protein, moisture, and oil content directly at the point of harvest or sale rather than waiting for centralized laboratory results.⁷

A comprehensive review by Jerome Workman Jr. and Will Wetzel highlighted how combined ATR-FTIR and NIR screening protocols are similarly expanding into environmental soil and water monitoring, extending NIR's traditional agricultural strengths into broader environmental surveillance applications with minimal sample preparation requirements.⁸

The Elusive Goal: Noninvasive Glucose Monitoring

No application illustrates both NIR's promise and its persistent limitations better than noninvasive blood glucose monitoring. Despite decades of research investment, no fully noninvasive optical glucose monitor has yet received FDA approval, largely because glucose is an inherently weak near-infrared absorber relative to water, lipids, proteins, and hemoglobin in tissue, making selective, interference-free detection extraordinarily difficult.⁹

“Noninvasive glucose monitoring has long been considered the 'holy grail' of diabetes management, offering the promise of accurate, painless, and convenient measurement without the need for blood sampling.”¹⁰

— Jerome Workman Jr., Spectroscopy Online

Recent academic efforts continue to chip away at the problem. Researchers at SRM Institute of Science and Technology in India developed a compact near-infrared spectrometer aimed at painless glucose measurement, while a team at Sharif University of Technology in Tehran presented a molar-absorptivity-based approach to blood glucose estimation that achieved performance comparable to conventional principal components regression models.¹⁰ Persistent obstacles, including calibration drift, biological interference from skin composition variability, and limited dataset sizes in feasibility studies, continue to separate promising laboratory demonstrations from clinically validated, FDA-cleared consumer devices.

Hyperspectral NIR Imaging Extends Beyond Point Measurements

Beyond single-point spectral acquisition, NIR chemical imaging is combining spectroscopy with spatial mapping to visualize component distribution across a sample surface rather than reporting a single averaged spectrum. Hyperspectral NIR imaging systems, which collect a full NIR spectrum at every pixel of a two-dimensional detector array, are increasingly used in pharmaceutical manufacturing to assess blend uniformity and active pharmaceutical ingredient distribution across tablet surfaces, and in food science to map moisture, fat, and adulterant distribution across heterogeneous samples such as ground meat or grain kernels.

The Fabry-Perot field-calibration advances described above are particularly relevant to hyperspectral imaging, since collecting a full spectral cube historically required far longer calibration and acquisition times than single-point NIR measurements. As near-real-time calibration and acquisition methods mature, hyperspectral NIR imaging is transitioning from a specialized laboratory technique into a viable at-line and potentially in-line process analytical tool, a trend closely related to broader process analytical technology adoption discussed elsewhere in this series.

Summary and Conclusions

Near-infrared spectroscopy's trajectory over the past several years reflects a broader pattern across analytical spectroscopy: miniaturization and chemometric sophistication are extending a mature technique into new physical form factors and application domains rather than replacing its underlying physics. Chip-scale MEMS and linear-variable-filter sensors are shrinking FT-NIR instrumentation from benchtop systems into embeddable components, while faster field calibration methods, such as the reconstruction-matrix approach patented by Hinalea Imaging, are removing practical deployment barriers for hyperspectral and NIR imaging systems in the field.

At the same time, the noninvasive glucose monitoring challenge illustrates that miniaturization and computational sophistication alone cannot overcome fundamental physical limitations, namely, glucose's weak and confounded NIR absorption signature amid a complex biological matrix. As Professor Ozaki's career-spanning contributions in overtone and combination-band theory demonstrate, continued advances in fundamental NIR spectral understanding remain just as essential to progress as instrument hardware innovation.

Across every application discussed here, from grain silos to pharmaceutical blend uniformity to the still-unsolved problem of noninvasive glucose sensing, the throughline is the same: NIR spectroscopy's future depends as much on chemometric and computational sophistication as on the optical hardware collecting the spectrum in the first place.

Future Outlook

Expect continued proliferation of chip-scale NIR sensors into consumer electronics, agricultural equipment, and point-of-need food safety testing as manufacturing costs decline and semiconductor fabrication techniques mature further. Competition between MEMS-interferometer and linear-variable-filter chip architectures is likely to continue, with cost, spectral range, and resolution trade-offs determining which approach dominates specific application niches.

In noninvasive glucose monitoring, near-term progress will likely come from multimodal sensor fusion, combining NIR with other optical or bioimpedance modalities, and from larger, more diverse clinical training datasets for machine-learning calibration models, rather than from NIR hardware improvements alone. More broadly, as two-dimensional correlation spectroscopy, anharmonic quantum chemical calculations, and deep-learning chemometric methods mature in parallel, the interpretability of NIR's inherently broad and overlapping spectral bands should continue to improve, narrowing the gap between NIR's practical convenience and the chemical specificity traditionally associated with mid-infrared and Raman techniques.

The competition between MEMS-interferometer and linear-variable-filter sensor architectures is also likely to spill over into wearable consumer health and wellness devices, building on the same underlying chip-scale manufacturing techniques that have already made spectral sensing inexpensive enough to embed in smartphones and handheld agricultural analyzers, even as regulatory clearance for medical-grade applications such as glucose monitoring continues to lag behind consumer-grade wellness use cases.

References

(1) Ozaki, Y. NIR Spectroscopy – What a Wonderful World! NIR news 2022, 33 (1–2), 10–17. DOI: 10.1177/09603360211067093.

(2) Spectroscopy Online Editorial Staff. A Life of Spectroscopic Exploration: An Interview with Professor Yukihiro Ozaki, Recipient of the 2025 Ellis R. Lippincott Award. Spectroscopy Online, 2025. https://www.spectroscopyonline.com/view/a-life-of-spectroscopic-exploration-an-interview-with-professor-yukihiro-ozaki-recipient-of-the-2025-ellis-r-lippincott-award (accessed 2026-08-12).

(3) Si-Ware Systems. Si-Ware Systems Introduces First Fully Integrated Chip-Sized Spectral Sensor for Industrial and Consumer Markets. Si-Ware Press Release. https://www.si-ware.com/our-news/si-ware-systems-introduces-first-fully-integrated-chip-sized-spectral-sensor-for-industrial-and-consumer-markets (accessed 2026-08-12).

(4) ams-OSRAM. AS7421 Hyperspectral NIR Sensor. Product Documentation. https://ams-osram.com/products/sensor-solutions/ambient-light-color-spectral-proximity-sensors/ams-as7421-hyperspectral-nir-sensor (accessed 2026-08-12).

(5) Shu, G.; Fong, A. Field Calibration for Near Real-Time Fabry Perot Spectral Measurements. U.S. Patent 11,867,615 B2, January 9, 2024 (assignee: Hinalea Imaging Corp.). https://patents.google.com/patent/US11867615.

(6) Peiris, K. H. S.; Bean, S. R.; Wu, X.; et al. Performance of a Handheld MicroNIR Instrument for Determining Protein Levels in Sorghum Grain Samples. Foods 2023, 12 (16), 3101. DOI: 10.3390/foods12163101.

(7) Chadalavada, K.; Anbazhagan, K.; Ndour, A.; et al. NIR Instruments and Prediction Methods for Rapid Access to Grain Protein Content in Multiple Cereals. Sensors 2022, 22 (10), 3710. DOI: 10.3390/s22103710.

(8) Workman, J., Jr.; Wetzel, W. Infrared and Near-Infrared Spectroscopy: Recent Advances and Environmental Applications (2022-2026). Spectroscopy Online, 2026. https://www.spectroscopyonline.com/view/infrared-and-near-infrared-spectroscopy-recent-advances-and-environmental-applications-2022-2026 (accessed 2026-08-12).

(9) Chen, Q.; Zhang, W.; Wu, W. VIS/NIR/FIR Spectroscopy for Blood Glucose Monitoring: A Review. AIP Adv. 2025, 15 (3), 030703. DOI: 10.1063/5.0239187.

(10) Workman, J., Jr. Noninvasive Glucose Monitoring Using Spectroscopic Methods. Spectroscopy Online, 2025. https://www.spectroscopyonline.com/view/noninvasive-glucose-monitoring-using-spectroscopic-methods (accessed 2026-08-12).