音声制御内視鏡ロボット（AESOP3000）の音声認識率向上に向けての発音方法の検討

概要

論文の詳細を見る
Objectives: AESOP3000 is a computer-driven robot capable of holding and positioning a laparoscope under the direct voice control of a surgeon. This device could be especially helpful when surgeon availability is low. Because we have occasionally experienced voice misrecognition in test situations, we conducted a study to determine whether or not voice training is an effective corrective measure for this problem.Methods: At voice card recording, we modified pronunciation and accent for a number of commands (AESOP, Back, Up, In, Down, Right, Left) in English-speaking mode, and carried out 26 of 63 laparoscopic maneuvers. Conventional voice cards were also used in the same series of operations for comparison. Results: With the stated modifications, voice misrecognition rates for most commands were significantly reduced as follows: AESOP: 17.5% to 9.2% (p<0.001), Back: 23.9% to 8.3% (p<0.001), Up: 18.5% to 6.7% (p<0.001), In: 9.4% to 8.0%, Down: 8.7% to 4.2% (p<0.05), Right: 4.3% to 3.4%, Left: 25.1% to 13.1% (p<0.001).Conclusions: While not 100% effective, pronunciation training is a valid means of reducing AESOP3000 voice misrecognition.