Detecting respiratory diseases using machine learning algorithms is important for early diagnosis and control of the spread of diseases, especially in epidemics or pandemics. The use of diagnostic methods based on the analysis of lung breathing recordings can be an effective tool for detecting signs of respiratory diseases such as influenza, COVID-19, pneumonia, and others. In this paper, we developed an approach for detecting respiratory diseases in lung breath recordings that uses a histogram-based gradient boosting classifier . To reduce time and resources, less complex algorithms were chosen, but the main goal is to obtain a one-dimensional array in audio data processing that can be used in standard machine learning algorithms and simple neural networks. The results of applying the proposed algorithm showed an accuracy of 93%, while the execution time was only 0.1 s, which allows for real-time solutions. The solution to this problem can greatly facilitate the work of medical staff by reducing the time spent examining people who do not actually have respiratory diseases.