Speech Intelligibility Transformation & Autocorrection

Universal software system for automated, customizable speech intelligibility enhancement for broadcast transmission and end listeners.

For people with reduced hearing ability, understanding speech while watching television can sometimes be strenuous or even impossible. The use of music, effects and background noise is sometimes a problem - not only for older or hearing impaired people.

In the BMBF-funded joint project SITA, solutions based on innovative software technologies have been researched, developed and evaluated that address all major factors for poor speech comprehension of audio content and comprehensively break down the barriers to intelligible speech. The project consortium consisted of the consortium leader RTW GmbH & Co. KG from Cologne, the Fraunhofer Institute for Digital Media Technology (IDMT), Oldenburg and ebee Engineering GmbH Dresden (formerly TechniSat Dresden GmbH).

The project goal was to develop a technology for the objective measurement of speech intelligibility and its automatic and target group-specific improvement. The system should not only be used in professional environments, such as radio and television stations, but should also give end customers the possibility to enjoy a better listening experience on the home television set.

Based on machine learning methods, the audio signals received by the system are analyzed in real-time and processed in a speech intelligibility model. This model takes into account the processing mechanisms that are essential for human speech perception and calculates an index that reflects the hearing effort perceived by humans.

As a result of the research project, software for use in the professional audio field was developed which numerically and graphically represents the results of the speech intelligibility analysis. Furthermore, a solution was developed that enables the user to automatically improve speech intelligibility in the home environment in a target group-specific manner.

The technologies developed in the research project enable not only radio and television stations, production companies, distributors and content providers to provide the end customer with a better listening experience, but also media service providers and manufacturers from the home entertainment and consumer electronics sectors can benefit from this.