Assessing normality of heart sound by matching pursuit residue with frequency-domain-based templates.

The healthiness of the heart is import to human for leading a healthy life. Some of the abnormal conditions of heart could be detected by the heart sound. For example, heart murmur is one of the conditions possibly resulting from the illness of heart valves. In this work, we propose a new template family used in the matching pursuit (MP) algorithm to assess the degree of normality of the heart sound. Not only the rating of normality is generated but also an intuitive visual feedback with abnormal interval highlighted is outputted. In this way, the condition like heart murmur is well-identified by inspecting the matching residue. Comparison with previous work which adopts different template family is also given. Experiment results indicate that the proposed family is more effective to matching the normal heart sound to achieve better assessing result.