Evaluation of a portable two‐microphone adaptive beamforming speech processor with cochlear implant patients

A two‐microphone noise reduction technique was tested with four cochlear implant patients. The noise reduction technique, known as adaptive beamforming (ABF), used signals from only two microphones—one behind each ear—to attenuate sounds not arriving from the direction directly in front of the patient. The algorithm was implemented in a portable digital signal processor, and was compared with a strategy in which the two microphone signals were simply added together (two‐microphone broadside strategy). Tests with the four patients were conducted in a soundproof booth with target speech arriving from in front of the patient and multitalker babble noise arriving at 90 deg to the left. Results at 0‐dB signal‐to‐noise level (S/N) showed large improvements in speech intelligibility for all patients, when compared to the two‐microphone broadside strategy. Precautions were taken to avoid cancellation of the target speech, and, accordingly, subjective tests showed no deterioration in performance for the adaptive beamformer in quiet. Physical measurement of the directional characteristics of the ABF was made with the microphones placed behind the ears of a KEMAR manikin and in the same acoustic environment as used with the patients. Results showed directional gain of approximately 10 dB when the angle of incidence for interfering noise was shifted more than 20 to 30 deg from directly in front of or behind the manikin. The effects of reverberation were explored by placing the manikin in different acoustical environments and observing the attenuation of the noise alone at various angles. A near anechoic environment allowed the noise to be attenuated by as much as 21 dB as the angle of incidence of the noise shifted away from directly in front of or behind the manikin. In a highly reverberant concrete stairwell, the adaptive beamformer was, however, unable to provide any directional gain beyond about 3 dB for any angle of incidence.