gms | German Medical Science

In daily practice, speech recognition scores in noise can differ with respect to measurement conditions such as room acoustics, noise source direction, or the position and head orientation of a test person. These parameters affect the signal-to-noise ratio (SNR) at the ears, and binaural cues used to segregate speech and noise signals to improve speech recognition in noise, also known as spatial release from masking [1]. Therefore, speech-recognition scores in noise measured under different conditions are not directly comparable. In particular, when measuring speech recognition in noise for diagnostic purposes or to evaluate hearing aid fittings, it is important to know to what extent these parameters affect the results. To investigate the effect of room acoustics and head orientation on typical spatial configurations of speech (S) and noise (N) signals, head movements and speech recognition thresholds (SRT) were measured in participants with normal-hearing in four different rooms for the loudspeaker configurations S0N0, S0N45, S0N90, S0N180 and S±45N∓45. Additionally, head orientations and positions were reproduced with an artificial head, the impact on SNR was measured and the effect on the SRT was calculated using a speech intelligibility model. The results confirm that different measurement conditions have an essential effect on speech-recognition scores in noise. The directions of speech and noise sources and thus the loudspeaker configuration has the largest effect on SRT due to spatial release from masking. Horizontal head rotations can additionally affect speech recognition scores for the configuration S0N180. Reverberation in rooms reduces the effect of spatial release from masking, which leads to a lower improvement of the SRT for spatially separated speech and noise signals compared to the configuration S0N0.