In non-invasive wearable spectral sensors, ensuring the stability of the contact state between the human body and the sensor is essential for optimal sensor performance. Human movements and changes in body posture can cause variations in muscle and skin tension, altering the contact state at the interface between the sensor and the skin, which results in shifts in the angle of incident light. Such changes affect the path of light propagation within the skin, subsequently influencing the intensity of the diffusely reflected light received by the sensor. To address this issue, this study developed a wearable near-infrared optical sensor specifically designed for non-invasive analysis of body components. The photosensitive part of the sensor features a ring-shaped design, which not only increases the area that receives light but also allows for the average of the signal at 360 degrees, effectively reducing the anisotropic interference caused by variations in the angle of incident light. Monte Carlo simulations were used to analyze the performance differences between point detectors and the ring-shaped detector under conditions where the light incident angles varied randomly within the ranges of 1.2°, 2.5°, 5°, 15°, and 45°, as well as in a monotonic increase from 0° to 5°. The results demonstrated that the ring-shaped detectors signal-to-noise ratio is significantly higher than that of the point detector—approximately ten times greater. Accordingly, its detection limits for glucose are also lower, one-tenth those of the point detector. The ring-shaped detector not only reduces the anisotropic interferences caused by random variations in the angle of incidence but also shows stronger common-mode characteristics in interference received at two different source-detector separations. Therefore, using a pair of concentric ring-shaped detectors and differentially canceling the signals can further suppress interferences caused by changes in the angle of incidence. In human trials, subjects used ring-shaped and point detectors to collect signals under changes in body posture while fasting. The experimental results indicate that the ring-shaped detectors, combined with the differential method, effectively suppress interference caused by changes in the human-sensor contact state due to body posture changes, with signal variation ranging between 0.000 5 and 0.001 a. u., thus meeting the non-invasive blood glucose detection requirements of 0.5 to 1 mmol·L-1. In summary, the dual-ring shaped wearable detector proposed in this paper exhibits outstanding high signal-to-noise ratio performance in human testing, demonstrating its broad potential for application in non-invasive human body component analysis.