Broad bandwidth helps listeners understand fragmented speech

Listening to a conversation in a noisy room or a crowded bar is no easy task. Why that’s the case is not fully clear. Recent studies have suggested that the main auditory difficulty arises not from isolating the target speech but from comprehending a version that has bits and pieces missing. Now Virginia Best and colleagues at Boston University have made a new observation: Systematically removing sound energy at high frequencies can make it harder to understand speech that’s partially masked by competing talkers.

Because speech is a highly redundant signal, robust information in one frequency band can compensate for lost information in another; the brain fills in the gaps (see the article by Emily Myers, Physics Today, April 2017, page 34 ). The moments during which the speech of interest dominates the acoustic mixture are known as glimpses. Best and her colleagues hypothesized that a loss of audibility at certain frequencies would hinder a listener’s ability to understand glimpsed speech more than intact speech.

Best and coworkers presented normal-hearing adults with audio of three sentences recited simultaneously. The researchers altered the recordings by filtering out segments that were overpowered by two of the talkers, leaving only the glimpses of the target sentence. (The frequencies and times at which the target speech dominates in three different recordings are shown in the graph; two of the recordings are in the sound clip below.) To test the effect of bandwidth on intelligibility, sound frequencies above a cutoff between 500 and 8000 Hz were systematically eliminated from the recordings. In each listening situation, Best and her colleagues measured speech intelligibility by scoring the listeners on the words they reported hearing. She found that as a greater proportion of the signal was blocked and glimpses were lost, high-frequency components (1000–8000 Hz) became increasingly important for making sense of what remained. Yet those high frequencies did not affect the intelligibility of intact, unmasked speech.

Hearing loss often starts at the highest frequencies and works its way down. The new findings suggest that those high frequencies are more important in noisy situations than in quiet ones, which may be an important consideration for improving the design of hearing aids. (V. Best et al., J. Acoust. Soc. Am. 146, 3215, 2019 .)