Cardiovascular diseases (CVD) or heart disorders are the leading cause of morbidity and mortality among men, women and people of most racial and ethnic groups in the world. CVDs currently pose the highest threat to human health around the world. Due to the cardiac dysfunction that affects the heart valves, blood flow is in turn affected, which can eventually lead to abnormal heart sound. It can therefore be adequately characterised by the phonocardiogram (PCG) signal, as it directly reflects the mechanical function of the human heart. The main issue is the non-linear character of phonocardiography signals and the wide range of distinguishable pathological PCG signals. The present work proposes a computer-aided technique based on PCG signals with the higher-order spectra analysis (HOS). The HOS method is used since it is well known as suitable tool for analysing non-linear bio-signals. A variety of PCG signals were used and converted to bispectrum 2-D images using the FFT direct method, then a HOS-based features extraction was carried out and the analysis was performed with the energetic ratio (ER) as benchmark parameter. It is shown that the distinctive bispectral pattern of the different PCG signals presented in this study implies that each phonocardiogram could be distinguishable by applying the graphical bispectral analysis. Moreover, the derived HOS-related features, such as bispectral magnitude, bispectral entropies and weighted center of the bispectrum, show good performance in assessing the severity of cardiac pathology in the heart sounds signals. It is found that these results showed that the variation of the extracted HOS features are sensible to the added murmur intensity in the PCG recordings. The results suggest that the information obtained from the use of HOS technique to analyse PCG signals can improve the diagnosis of heart disorders and help clinicians differentiate patients according to the pathological severity of their heart state.