Evaluation of a hearing compensation algorithm

David V. Anderson, Richard W. Harris, and Douglas M. Chabries Brigham Young University, Provo, Utah ABSTRACT A new hearing compensation algorithm based on a homomorphic multiplicative AGC (automatic gain control) is evaluated and compared against commercially available digitally programmable analog hearing aids. Both quantitative (speech recognition and threshold) and qualitative tests were used in the evaluation. The new algorithm is shown to have made signi cant progress in restoring normal or near normal hearing for hearing impaired individuals. 1. Hearing Compensation Hearing loss may occur for any of a number of reasons including noise exposure, old age, disease, and trauma; or an individual may be born with a hearing disorder. The types of hearing loss that result are also quite varied and include damaged inner or outer hairs cells within the cochlea, damaged or missing middle ear bones or tympanic membrane, and defective neural pathways. The methods used in compensating for hearing loss, on the other hand, are few and have usually involved only simple ampli cation, or ampli cation with high frequency emphasis. Recently, some hearing aids have begun using loudness compression to more accurately compensate for some hearing losses. A hearing compensation algorithm is evaluated which provides an accurate way to compensate for several types of hearing losses. The new hearing compensation algorithm is based on a homomorphic AGC (automatic gain control). It lters sound into multiple bands with an approximate bandwidth of one{third octave each. The sound in each of these bands is modi ed by an AGC of the form shown in gure 1. The outputs of all the AGCs are then recombined by simple summation. This approach shows signi cant performance gains based on intelligibility scores, SRT's (which are explained later), and also restores intelligibily in noise corrupted environments to near normal performance levels. The hearing aid is based on a process model of the ear and provides precise multichannel loudness compression tailored the an individual's hearing loss [1]. The homomorphic AGC hearing compensation algorithm was implemented on a real-time DSP board with a 16 bit ADC and DAC and a sample rate of 21.333 kHz. The sound was ltered into 12 one-third octave bands, the loudness of each band was compressed in such a way as to greatly increase the intensity of soft sounds but not loud sounds, and the bands were each recombined and the sound was played through a standard hearing aid receiver (speaker). This setup is hereafter referred to as the \new aid." The \new aid" is compared against two other hearing aids that were deemed to be the best digitally controlled analog behind{ the{ear aids commercially available. These two hearing aids s (t) s (t)