In audio mixes, different speakers at the same time are not unusual – but this sometimes compromises speech intelligibility. Fraunhofer IDMT in Oldenburg has a solution. Voice filtering, which produces the acoustic fingerprint of a speaker within a few seconds so that it can be extracted later from a mix of several speakers, enables the reliable recognition and separation of different speakers. On the basis of our speaker authentication technology the different audio signals are filtered out of the audio mix in order to separate the speakers.
Voice filtering can also be used to remove background noise or interference from a mix. This increases both speech intelligibility and sound quality.
Voice filtering for people with hearing impairments
Especially people with hearing impairments can have difficulty concentrating on and listening to one person in a group. This is also known as the cocktail party effect. Here, the voice filter technology’s intelligent algorithms can help to filter out one person from the complicated mix and in this way improve speech intelligibility.
Voice filtering in telecommunications
The technology not only helps people with hearing impairments to focus on the specific person speaking but is also used in telecommunications or in the transmission of speech over the internet, i.e. VoIP (Voice over Internet Protocol). Minimising interference raises the intelligibility of spoken language, for example in online meetings, video conferences or call centres.
So that only a specific person can issue voice commands to a system, voice filtering is helpful in production. Here, too, ambient noise and other speakers are filtered out. This is particularly advantageous in noisy environments in order, for example, to increase the accuracy of speech recognition. The voice filtering technology can also be used in the security sector.