Audio scene analysis for hearing aids
Marie A. Roch, Tong Huang, Richard R. Hurtig · The Journal of the Acoustical Society of America · 2004
It is well known that simple amplification cannot help many hearing-impaired listeners, and numerous signal enhancement algorithms have been proposed for digital hearing aids. In many cases, algorithms are most effective in specific environmental or source conditions. If one can properly detect components of the auditory scene, it is possible to dynamically apply enhancement algorithms which are appropriate for a given situation. This work illustrates this principle by describing a cohort detection scheme which serves as a control system for a frequency-domain compression algorithm. The compression algorithm preserves formant ratios and thus enhances speech understanding for individuals with severe sensorineural hearing loss in the 2–3-kHz range. By detecting speakers from broad cohorts (e.g., male, female), it is possible to adjust the compression ratio dynamically based upon characteristics of the auditory scene, resulting in a more appropriate enhancement than one based upon a static ratio. Cohort decisions are derived from the likelihood scores of a Gaussian mixture model classifier.