Examination of Balance Adjustment Method Between Voice and BGM in TV Viewing

As the elderly age, their hearing deteriorates and it becomes difficult to hear the sounds of daily life. Especially, “TV sound” is hard to hear for one in two elderly people. The cause is that the voice is drowned out by the BGM. In this research, we focus on the volume balance between voice and BGM in TV sound, and propose a volume balance adjustment method using sound source separation technology as a method to adjust these appropriately. In addition, the effectiveness of the proposed method will be evaluated through subjective evaluation. In order to improve your hearing of TV sounds, you need to emphasize them. Therefore, in this research, we propose a method to emphasize the sound by separating the TV sound into voice and BGM by the sound source separation technology, suppressing the gain of BGM, and then reintegrating it. In this study, Spleeter is used. Spleeter is a sound source separation software that uses supervised deep learning. It is mainly used to separate songs into parts, and the input music data can be divided into parts (Example: Vocal/Accompaniment). In the experiment, we used a mixture of voice and BGM as the sound of the TV. (We have prepared two types of voice, “Natural voice” and “Whispering voice”.) This simulated data is separated by Spleeter, and the gain of the BGM after separation is suppressed and mixed again. Eight male subjects in their twenties will be asked to hear the sound before and after processing to evaluate whether the ease of hearing the voice can be improved. As a result, it was found that increasing the ratio of voice improves the ease of hearing the voice. However, it was also found that the distortion generated in the process of sound source separation also affects the sound quality. Therefore, it can be said that it is necessary to improve the accuracy of sound source separation in order to further enhance the effect. In this study, we proposed a method to adjust the volume balance between voice and BGM to an appropriate level using sound source separation technology. In the future, we would like to consider ways to further improve hearing by improving the accuracy of sound source separation.