Predicting speech-in-speech recognition: Short-term audibility and spatial separation.

Area of Science:

• Auditory science
• Speech perception
• Psychoacoustics

Background:

• Variability in speech-in-speech recognition is a key challenge.
• Energetic and informational masking (IM) influence recognition, but IM is hard to quantify with dynamic maskers.

Purpose of the Study:

• To investigate how short-term audibility and spatial separation affect speech recognition in two-talker masking.
• To quantify the impact of spatial cues on informational masking.

Main Methods:

• Thirty young adults with normal hearing (NH) performed sentence recognition tasks.
• Speech reception thresholds and keyword recognition were measured under varying spatial conditions.
• A glimpsing model quantified short-term audibility for keywords.

Main Results:

• Speech recognition correlated with the proportion of audible keyword glimpses, irrespective of overall signal-to-noise ratio (SNR).
• Greater short-term audibility was needed for colocated speech compared to spatially separated speech.
• Spatial separation enhanced the benefit of increased target audibility on keyword recognition.

Conclusions:

• Spatial cues improve the efficiency of glimpsing in competing speech for individuals with normal hearing.
• The study provides a quantitative framework for estimating informational masking in different spatial configurations.