Breakthrough technology from Cambridge audio software specialist AudioTelligence is set to help solve the ‘cocktail party problem’.
80% of adults aged between 40 and 64 find it hard to follow conversations in noise – with over 200 million people worldwide looking for a solution, according to the company’s research. But AudioTelligence’s new Aiso for Hearing product has been demonstrated to improve speech understanding in noisy environments from 5% to 98%.
Current assistive listening technologies, based on beamforming and noise suppression, provide only a partial solution to the cocktail party problem. When hearing aids are used in noisy environments, they tend to amplify all the sounds – including the background noise – so the user still finds it difficult to follow the conversation.
Aiso for Hearing adopts a different approach – instead of amplifying all sounds from a particular direction, Aiso separates the speech signals from the surrounding noise. This allows a user to hear individual voices clearly, with the technology giving an improvement in the speech-to-noise ratio (SNR) of up to 30 decibels (dBs) – compared with improvements ranging between only 2-4 dBs for the directive beamformers typically found in hearing aids.
Aiso for Hearing is a software solution which can be integrated into multi-microphone devices – including smartphones and hearing solution products – to give them enhanced assistive listening capabilities. Aiso for Hearing is unique in using a combination of patented blind source separation and noise suppression for a two-pronged approach to the cocktail party problem. The technology first analyses the acoustic scene, separating sound sources and isolating the voices of interest. It then uses noise suppression techniques to reduce the background noise. AudioTelligence’s expertise as the world leader in blind source separation from multi-microphone devices has enabled the company to develop both technologies with very low latency (5ms). Low latency is crucial for assistive listening solutions which need to work in real time, and which must avoid ‘lip sync’ issues.
Ken Roberts, CEO of AudioTelligence, said: "Our alternative very low latency approach separates voices from noise, improving speech intelligibility, whereas other solutions tend to just amplify sound from a particular direction. At the same time, Aiso for Hearing suppresses background noise, minimising the effort needed to follow a conversation and reducing listening fatigue. And if there are multiple conversations happening around them, the user can pick which one they wish to focus on, eliminating the unwanted speech.
"Our technology portfolio is the result of more than 12 years of research into the most effective approaches for extracting clean speech signals in complex acoustic scenes. This has enabled us to develop assistive audio technology to solve the cocktail party problem and help 200 million people worldwide to join the conversation."
