Real-time optimization of speech intelligibility
The speech reproduction of communication systems is often subject to superimposed reverberation and background noise. With AdaptDRC, the Fraunhofer IDMT offers a patented software solution which automatically improves speech intelligibility in real time, even in hearing situations with unknown, variable interference noise.
Near-end listening enhancement – signal adaptation for interference noise on receiver side
The acoustic situation on the receiver side is detected by means of a microphone – e.g. on the rear side of cell phones – and the intelligibility of the speech signal in the current hearing situation is analyzed. On the basis of this analysis, model-based signal processing methods optimize speech intelligibility in real time. By taking into account the latest findings from hearing research, AdaptDRC achieves a high level of hearing comfort even for narrow-band signals and people suffering from hearing loss.
Hearing model-based signal processing
AdaptDRC uses human hearing perception models to evaluate and improve the intelligibility of a speech signal. The algorithm continuously analyzes the current intelligibility based on the speech intelligibility index SII. Signal processing is activated when intelligibility decreases, e.g. due to an increase in interference noise. Individual frequency bands are selectively boosted and the dynamic range of the signal compressed at the same time. This makes it possible to improve speech intelligibility by 30-80 percent without increasing the volume. Scientific studies with test persons with both normal and impaired hearing have shown that persons with hearing loss also benefit from signal processing and are able to understand speech signals more easily and with less hearing effort. The fact that the signal is processed only if speech intelligibility is impaired contributes to hearing comfort and naturally sounding speech reproduction.
AdaptDRC makes it possible to improve speech intelligibility in different technical systems and with different bandwidths – from conventional telephone transmission through to full bandwidths in high-quality multimedia applications. Due to the dynamic signal compression used, speech intelligibility can be improved even in difficult acoustic environments. The algorithm is available as a platform-independent C implementation and can be adapted to application-specific scenarios and technical conditions.