LEARNING APPARATUS, LEARNING METHOD, PROGRAM, LEARNT MODEL AND LIP READING APPARATUS

摘要

To perform practical lip reading estimation even for an infrared video for which teacher data are difficult to obtain.SOLUTION: A learning apparatus for learning a model of a neural network used for estimating an utterance content from movement of lips of a person imaged in an infrared video includes: an RGB video model storage unit 23 that stores an RGB video model configured by learning weighting coefficients of the neural network by a learning unit 22 using the RGB video including a video of the lips and the utterance content corresponding thereto as teacher data; an infrared corpus input unit 10 that inputs the infrared image including the video of the lips and the utterance content corresponding thereto as the teacher data; and a transfer learning unit 11 that reads the RGB video model from the RGB video model storage unit 23, performs transfer learning of the read learnt model using the teacher data input by the infrared corpus input unit 10, and generates an infrared video model 12.SELECTED DRAWING: Figure 1