Stanford University

News Service



Dawn Levy, News Service (650) 725-1944; e-mail:

Microphone array necklace aids the deaf in discerning speech

Almost totally deaf and reliant on lip reading since her 20s, Sherry Cramer couldn't believe her ears in 1994 when she first wore the microphone array necklace that electrical engineering Professor Bernard Widrow and his students had designed. Listening to a CD, she could hear every note of a Rachmaninoff piano concerto as the necklace received and transmitted sound in magnetic form to her behind-the-ear hearing aid.

A childhood student of flute and piano, Cramer recalled her response during an interview that was videotaped in the mid-1990s: "My God, I can hear!" In her 30s at the time of the interview, she suffered from a deterioration in hearing that continues to this day. "I was shocked. I remember walking around and wanting to hear different sounds what's that? What's that? What's that?"

Widrow let her take the necklace home, and the hearing world opened up like a desert flower. Cramer was able to go to a movie for the first time in a decade and follow the plot, even when the characters were not facing her, enabling her to read their lips. Once again she could listen to the radio, which she had abandoned in her teens. An accounting manager, she began to interact with co-workers from whom she had shied away, and even gave presentations at which she answered questions something that had been virtually impossible before. Within the year, she was promoted to chief financial officer for a company of more than 300 employees. And when she went to a noisy restaurant to celebrate, she was able to hear not only her dinner companion but also the people at the next table. "I was eavesdropping for the first time in my life!" she recalled.

Since Cramer, hundreds have used Widrow's necklace. Widrow demonstrated his latest version, the Directional HEaring ARray (D-HEAR), June 5 and 6 at a Chicago meeting of the Acoustical Society of America, at which he delivered the society's distinguished lecture.


The original device was a somewhat clunky three-piece prototype that later evolved into a more compact unit that could be worn around the neck. In its present incarnation, six tiny microphones and signal-processing electronics are housed in a lightweight and elegant black plastic V-shaped casing. It looks more like a boomerang-shaped necklace than an electronic device.

Users of the hearing-aid necklace demonstrate dramatic improvements in speech discernment. According to a Widrow study funded by the National Institutes of Health and completed in 1999, all nine patients who were administered a hearing-in-noise test were able to repeat significantly more words when wearing the microphone array necklace and a hearing aid than when wearing a hearing aid alone. The first patient, for example, was able to discern 25 percent of words correctly when wearing a hearing aid but 95 percent when wearing a hearing aid and the microphone array necklace.

The necklace has been designed to aid those with severe to profound hearing loss. Widrow estimates that as many as 2 million people experience this degree of hearing impairment in the United States. As hearing-aid manufacturers prefer to develop products for the larger market those with moderate hearing loss no company currently markets the necklace array. It was originally manufactured by Cardinal Research LLC of Stanford, Calif., of which Widrow is chairman, and subsequently by Starkey Laboratories of Eden Prairie, Minn.

Microphone array technology is crucial to improved performance of hearing aids. Most hearing aids feature only one microphone, but amplification from a lone microphone has limitations: Noise is amplified as well as signal. Cacophony can result, especially in noisy rooms, making it difficult for hearing-impaired people to understand amplified sound. "There is a big difference between hearing speech and understanding speech," Widrow says.

So how does the necklace, patented by Widrow and his former graduate students Michael A. Lehr and Stephen W. Mims, boost signal but mute noise? The user wears both the necklace and a hearing aid. Sound waves enter from a 60-degree-wide cone-shaped space in front of the user. In a noisy place, the user orients his or her body toward the speaker and surrounding sound is minimized. Microphones in the necklace pick up the sound and transmit it to signal-processing chips that give different weights to input sounds from the various microphones. The weights have been determined through a computer design procedure based on the LMS (least mean squared) algorithm. The result? The microphone array is able to home in on the desired signal and reduce echoes and other undesirable auditory effects while increasing clarity of the dominant signal. The optimized signal is then amplified and sent through a conducting neckloop, which wirelessly transmits a magnetic signal to the telecoil in the user's hearing aid. Hearing aids commonly feature the telecoil to facilitate use of a telephone by a hearing-impaired person.

Bernard Widrow

An eminent inventor, Widrow in May was awarded the 2001 Benjamin Franklin Medal in Engineering, a prestigious award whose past winners include Thomas Alva Edison, Guglielmo Marconi, Orville Wright, Alexander Graham Bell and Claude Elwood Shannon. Inducted into the National Academy of Engineering in 1995, Widrow is also a Life Fellow of the Institute of Electrical and Electronics Engineers (IEEE) and recipient of IEEE's Centennial Medal, Alexander Graham Bell Medal and Neural Networks Pioneer Medal. At Stanford since 1959, he was educated at MIT, where he received his bachelor's (1951), master's (1953) and doctoral (1959) degrees.

He is a trailblazer in the fields of adaptive signal processing and neural networks. In adaptive signal processing, signal-processing systems improve their performance through "learning by experience." Adaptive signal processing employs Widrow's LMS algorithm to reduce noise in a signal. It has been employed in computer-data transmission, military surveillance, adult and fetal electrocardiography, control systems to stop vibration and cancel noise, and adaptive control systems in aircraft and chemical plants. Neural networks, on the other hand, are trainable, engineered systems whose designs are inspired by the brain's neurons and their interconnections. Applications include speech recognition, pattern recognition and control systems.

Kathleen O'Toole contributed to this report.


By Dawn Levy

© Stanford University. All Rights Reserved. Stanford, CA 94305. (650) 723-2300. Terms of Use  |  Copyright Complaints