Speech Processing Algorithms for Older Listeners with Hearing Loss

Age-related hearing loss is a highly prevalent chronic health condition that severely limits the older person’s ability to communicate during social, occupational, and other personal interactions and in turn, affects quality of life.  Hearing aids are the principal form of remediation, but only 20% of elderly people with significant hearing loss obtain them.  One of the primary reasons for limited use of hearing aids is that they do not provide sufficient benefit in noise or for understanding the speech of certain talkers, such as those who speak at a fast rate.  The current project aims to address these two limitations of hearing aids through a systematic assessment of two types of signal processing algorithms that potentially can benefit older hearing-impaired listeners.  One type of signal processing algorithm, developed by Dr. Carol Espy-Wilson, seeks to separate the target speech signal from a competing speech signal at key intervals in the speech stream.  The second type of signal processing algorithm, based on research of Dr. Sandra Gordon-Salant, attempts to slow certain brief segments of the speech signal (i.e., consonants) that are difficult for older hearing-impaired listeners to process, while at the same time attempts to accelerate relatively long segments of the speech signal (i.e., vowels and sonorants) that are highly redundant for older hearing-impaired listeners.  Experiments are planned to assess the benefits of each of these signal processing algorithms for older hearing-impaired listeners.  Preliminary studies have compared the difficulty of younger and older listeners with normal hearing for understanding natural fast-rate speech in comparison to natural-rate speech and time-compressed speech.  These pilot studies have also examined the acoustic characteristics of natural fast speech.

Related Publications and Presentations

  • Zion, D., Espy-Wilson, C., Gordon-Salant, S. Recognition of natural-rate, time-compressed, and natural fast-rate sentences by younger and older listeners. 5th International and Interdisciplinary Research Conference on Aging and Speech Communication. View Link
  • Gordon-Salant, S., Zion, D., & Epsy-Wilson, C. (2014). Recognition of time-compressed speech does not predict recognition of natural-fast speech sentences by older listeners. Journal of the Acoustical Society of America -- Express Letters, 136, EL 268-274. View Article

This research is supported by an NSF-supported grant to the ADVANCE Program, UMD.