Relationship Between Working Memory, Compression, and Beamformers in Ideal Conditions.

Objectives: Previous research has shown that speech recognition with different wide dynamic range compression (WDRC) time-constants (fast-acting or Fast and slow-acting or Slow) is associated with individual working memory ability, especially in adverse listening conditions. Until recently, much of this research has been limited to omnidirectional hearing aid settings and colocated speech and noise, whereas most hearing aids are fit with directional processing that may improve the listening environment in spatially separated conditions and interact with WDRC processing. The primary objective of this study was to determine whether there is an association between individual working memory ability and speech recognition in noise with different WDRC time-constants, with and without microphone directionality (binaural beamformer or Beam versus omnidirectional or Omni) in a spatial condition ideal for the beamformer (speech at 0 , noise at 180 ). The hypothesis was that the relationship between speech recognition ability and different WDRC time-constants would depend on working memory in the Omni mode, whereas the relationship would diminish in the Beam mode. The study also examined whether this relationship is different from the effects of working memory on speech recognition with WDRC time-constants previously studied in colocated conditions.

Design: Twenty-one listeners with bilateral mild to moderately severe sensorineural hearing loss repeated low-context sentences mixed with four-talker babble, presented across 0 to 10 dB signal to noise ratio (SNR) in colocated (0 ) and spatially separated (180 ) conditions. A wearable hearing aid customized to the listener's hearing level was used to present four signal processing combinations which combined microphone mode (Beam or Omni) and WDRC time-constants (Fast or Slow). Individual working memory ability was measured using the reading span test. A signal distortion metric was used to quantify cumulative temporal envelope distortion from background noise and the hearing aid processing for each listener. In a secondary analysis, the role of working memory in the relationship between cumulative signal distortion and speech recognition was examined in the spatially separated condition.

Results: Signal distortion was greater with Fast WDRC compared with Slow WDRC, regardless of the microphone mode or spatial condition. As expected, Beam reduced signal distortion and improved speech recognition over Omni, especially at poorer SNRs. Contrary to the hypothesis, speech recognition with different WDRC time-constants did not depend on working memory in Beam or Omni (in the spatially separated condition). However, there was a significant interaction between working memory and cumulative signal distortion, such that speech recognition increased at a faster rate with lower distortion for an individual with better working memory. In Omni, the effect of working memory on speech recognition in different spatial conditions (colocated versus spatially separated) was inconclusive.

Conclusions: The findings highlight the benefit of binaural beamformers for all listeners, especially at poorer signal to noise ratios for target speech from the front and noise behind the listener. Individuals with better working memory are more likely to benefit from reduced signal distortions than individuals with poorer working memory in these conditions. There was no clear evidence for benefit or detriment in speech recognition with Fast versus Slow WDRC, regardless of individual working memory.