Speech Intelligibility in Radio Broadcasts : A Case Study Using Dynamic Range Control and Blind Source Separation

Abstract: Creating the optimal balance between dialogue level and ambient sound is extremely important in media productions. This process is however inherently difficult due to that people’s requirements and preferences are not uniform. Speech intelligibility is affected by a multitude of factors, such as hearing impairments, audio quality and listening equipment. Recent EU directives on accessibility calls for improved audio clarity features for broadcast content. To accommodate these requirements, the broadcast industry needs to develop functionality for enhanced dialogue clarity and, optimally, put listeners in control of these features. Many speech enhancement techniques exist, this paper uses Sveriges Radio as a case study to evaluate several of these methods. A study on enhancing speech intelligibility through the use of dynamic range control and blind source separation is presented and results show that both methods can have a positive impact. Dynamic range control proves efficient in increasing intelligibility by reducing dynamic variations. It is also well suited to implement in an existing two-channel infrastructure, common in the radio industry, due to being included in novel audio codecs. Blind source separation is found to best be used in moderation due to the risk of audio quality degradation, and is primarily suited for prerecorded material on account of the processing time needed.