Accurate assessment of metabolic rate is crucial for predicting human thermal comfort. However, many existing instruments are bulky, invasive, and inconvenient. Additionally, wearable metabolic rate measurements have faced long-standing challenges due to a lack of a comprehensive theoretical model with easy-to-measure parameters. This paper optimizes our previously proposed theoretical approach for wearable metabolic rate measurement that incorporates heart rate, whole-body unit heat loss, skin resistance, and body muscle percentage. We have improved the evaluation approach for whole-body unit heat loss by introducing a newly designed convective heat loss coefficient from various body segments. Regarding the wearable sensor design, a heat loss measuring system with serpentine channel was verified as the most suitable design following calibration experiments. We also integrated a modified square-wave-based skin resistance design, resulting in an upgraded integrated metabolic sensor. To validate our sensor and model performance, we conducted experiments with ten subjects of both genders at three temperature conditions (22, 25 and 28 °C) and four levels of activities (rest, tiptoe heel, slight walk, and moderate walk). First, our results demonstrate a strong positive relationship between the metabolic rate obtained by our optimized sensor and Quark CPET instrument, with high coefficients of determination (highest:0.97 from the male group, 0.98 from the female group). Our approach achieved at least 96 % accuracy and less than 2.5 % uncertainty for both groups. Second and increasingly, we found that only one metabolic rate model was required across combined temperature conditions when the person was fixed. Third, the positive relationship was further improved by incorporating whole-body unit heat loss. Finally, we obtained information from different segments on skin temperatures. In summary, our study provides a powerful methodology for predicting and evaluating human metabolic rate through an innovative wearable approach. Our approach and sensor have the potential to substantially contribute to energy-saving efforts in smart buildings.