Voice analysis for detection of hoarseness due to a local anesthetic procedure

This paper employs voice analysis techniques to detect hoarseness occurring as a result of administration of local anesthetic for shoulder surgery. The patients in our study developed transient vocal cord dysfunction; however surgery and anesthesia can also cause permanent vocal cord paralysis, which is a significant clinical problem. Early detection is important to optimize the chance of repair, and to avoid complications associated with an impaired swallow. We have developed an algorithm to detect altered vocal cord function, which is based on wavelet packet analysis (WPA) and support vector machines (SVM). This method is compared to the hoarseness diagram method (HDm), which was reported as an objective voice quality evaluation approach and can be used for pathological voice discrimination. We believe that our algorithm is unique in that samples were obtained from the same patients before and after their surgery. Our experiments using voice signals recorded from subjects before and after the procedure show a high classification accuracy, whereas HDm fails in the detection of a hoarse voice.