Application of the LMS adaptive filter to improve speech communication in the presence of noise

A major complaint of individuals with normal hearing and hearing impairment is a reduced ability to understand speech in a noisy environment. An adaptive LMS filter was employed to process speech in signal-to-noise ratios (S/N) varying from -8 to +12 dB. The filter configuration is commonly called noise cancellation. Intelligibility measures were obtained on normal and hearing impaired individuals using CID-W22 word lists. For the normal subjects mean intelligibility scores unprocessed were near zero for the S/N ratios -8, -4, and 0 dB and increased from 30 to 50% for the same S/N ratios when processed. For S/N ratios of +4 through +12, mean intelligibility scores prior to filtering ranged from 24 to 46%; after filtering, the scores improved from 67 to 75% i.e., within 7% of the nolse-free sound-field recordings. Preliminary data for the hearing impaired showed the same pattern with their improvement also approaching the scores obtained with no noise present in the sound field.