Studies on Development of Work Journal Supporting Systems with Speech Recognition

摘要

The farm journal recording system by speech recognition can decrease the burden of workers and help eliminate recording misses which may occur during the process in the field farm. This system consists of an IC recorder, a speech recognition system and work contents analysis divisions. In order to improve the ability of speech recognition software that converts voice to text, we researched into the compensation algorithm for the errors occurring after the voice is converted to text. In order to compensate for the conversion errors, a work dictionary is used to calculate the voice pattern, the character compatibility, suitable work period and work frequency to acquire the best possible work item. Because past records are used, the more the records are accumulated, the more the accuracy of the system is improved. At this stage, as enough data was available, this system could compensate for the conversion errors with high efficiency.