The method of detecting acetone levels in breath presents a promising approach for monitoring diabetes mellitus (DM). Consequently, the detection of acetone in exhaled breath is garnering significant attention. However, using ultraviolet differential optical absorption spectroscopy (UV-DOAS) for detection of the exhaled acetone has rarely been proposed due to the complex composition of exhaled gases and the baseline drift caused by the acetone absorption feature. In this study, we present an optical sensor based on an improved UV-DOAS and spectral upgrading, enabling the detection of exhaled acetone in the sub-200 nm wavelength band for the first time. Firstly, the overall fitting process in the UV-DOAS was improved to segmental fitting to address the issue of baseline drift, resulting in a standard differential absorption spectrum for acetone. Secondly, a spectral upscaling concentration inversion method based on wavelet coefficient matrix is proposed. This helps effectively handle spectral overlaps among oxygen (O