Snoring classification serves as a crucial tool for physicians in prescreening obstructive sleep apnea-hypopnea syndrome (OSAHS) and related sleep disorders. Current research on pathological snoring classification based on time-frequency domain features of audio signals faces limitations in classification accuracy, failing to meet clinical demands. To address this, we propose a high-dimensional snoring signal classification method utilizing feature texture (FT) images. The approach commences with data preprocessing, introducing a continuous respiratory event quality factor to determine the appropriate range of continuous respiratory signals. Subsequently, 51 time-frequency domain features are extracted and refined through the ReliefF algorithm for optimal feature selection and combination. These feature sets undergo normalization and grayscale conversion to construct feature texture grayscale images. Horizontal and vertical gradient matrices are derived from these images, which are then concatenated across RGB channels to form FT images. Experiments employ a convolutional neural network (CNN) to classify FT images derived from overnight respiratory sound signals of healthy subjects and OSAHS patients. Classification tests encompass breathing sounds, normal snoring, and OSAHS pathological snoring, revealing that the FT image approach achieves a classification accuracy of 97.2%, offering a promising foundation for subsequent sleep disorder management and research endeavors.