In this paper, combined with the continuous, reliable and low-cost real-time detection requirements of respiratory dust concentration, a spectral application technology innovation is realized, and the detection system with respiratory dust based on photo-acoustic spectroscopy is proposed. The center wavelength of the spectrum of the low-power diode laser is 403.56 nm and the corresponding effective absorption cross-section of NO2 is 5.948 5×10-19 cm2·mole-1. In addition, the resonance frequency of 1.35 kHz was obtained by using frequency scanning fitting. The influence analysis of photo-acoustic cell structure is carried out. The conclusion is drawn that the length parameter of the photo-acoustic cell has little influence on the background noise but great influence on the laser signal and the inner diameter parameter has some influence on the background noise but little influence on the background noise. Considering the influence of quality factors, processing conditions, service occasions and properties of the objects to be tested, length parameters of 120 mm and inner diameter parameters of 8 mm are selected. Besides, based on the buffer cavity structure with a length of 60 mm and an inner diameter of 25 mm, the influence of buffer partition on system stability is analyzed. The background noise is reduced and the signal fluctuation isoptimized from (2.83±0.11) to (1.26±0.03) μv. The specific absorption coefficient of NO2 with 195.28 Mm-1·(mg·m-3)-1 is analyzed. The system is calibrated using NO2 gas absorption at 405 nm. The fitting slope is 0.043 68 μv/Mm-1, the correlation coefficient is 0.998, and the pool constant is 300.24 Pa·cm·w-1. At the same time, the lower limit of the detected concentration and the absorption coefficient are 2.30 μg·m-3 and 0.448 Mm-1. The influence of absorption coefficient of respirable dust on polystyrene based on standard microspheres as aerosol generator is analyzed. What’s more, the absorption coefficients of particles with different concentration and different diameters at the same concentration are measured. The absorption coefficient of respirable dust is directly proportional to the number concentration. The slope after linear fitting is 10.598±0.641 96, and the correlation coefficient is 0.993. The variance of the absorption coefficient curve is between 3～4 Mm-1, and the absorption coefficient is affected by particles of different particle sizes. At the same time, the absorption coefficient increases with the increase of particle size. The NO2 measurements have been carried out in the ambient atmosphere. The filter membrane with 0.2 μm has been used to remove dust interference. The experimental results show that the concentration of NO2 in the atmosphere is 16.4~61.6 μg·m-3, and the average concentration is 41.1 μg·m-3. In order to verify the accuracy of the measurement system, the long-path differential absorption spectrum system developed by our group is compared. The test results show that there is a good correlation between the concentration of NO2 measured by the photo-acoustic spectroscopy system and the LP-DOAS system. The slope after linear fitting is 1.011 78±0.040 13, and the correlation coefficient is 0.947 81. The respirable dust in the ambient atmosphere is measured. The filter membrane with 5 μm is selected to filter the ambient atmosphere. The “NO2+5 μm dust” and “NO2+0.2 μm dust” are measured. The change trend of respirable dust is obtained, which can satisfy the real-time measurement of respirable dust absorption coefficient under natural suspension state.