The research group of Professor Shoji Makino of the Life Science Center for Survival Research, University of Tsukuba and Associate Professor Junki Ono of the Informatics Principles Research Institute of the National Institute of Informatics is one person from the overlapping voices of multiple people recorded by multiple devices. We have developed a voice signal processing technology that separates one voice and enables simultaneous recognition of multiple voices.It is expected to be used for automatic creation of meeting minutes.
According to the University of Tsukuba, the research group has developed a new signal processing technology that synchronizes multiple unsynchronized recorded signals recorded on different devices after recording.By combining this with the high-speed blind separation technology developed by the Ono Laboratory of the National Institute of Informatics, speech recognition is applied and recognized after separating conversations in which the voices of multiple speakers are mixed into individual voices. We have succeeded in greatly improving the performance.
By applying this technology, it will be possible to automatically create minutes if the conference participants are recording the conference content on their individual smartphones.
The voice recognition function has improved significantly in recent years, but when it is used for conversations and conferences in which multiple people's words overlap, the problem is that the function deteriorates.
The National Institute of Informatics is also researching how to implement this technology in systems on the web.The research results were presented at an international conference in the field of speech held in San Francisco, California, USA from September 9.
