Chip Imitates Ear to Clean up Cell Calls

If instead of writing this story I were telling it to you on my cell phone, while rushing through San Francisco International Airport past crying babies and the boarding announcements for Delta Flight 7867 to Seoul, you probably wouldn't understand half of what I'm about to say.

Cell phones frequently sound terrible, and often it's not that you can't hear anything. The problem is that you can hear too much. Unintelligible background noises compete with the intelligent voices trying to communicate. Lloyd Watts thinks that to solve that problem, we need to make cell phones more like us.

"Here's the insight that allowed us to build a product out of this little bit of neuroscience that we were doing. We said, 'If we could put a second microphone onto the cell phone, we could turn the cell phone into a creature that has two ears," Watts told an audience at the Hot Chips conference at Stanford last week.

Watts is CTO of Audience, which is based in Mountain View, California, and in February introduced a custom chip for cell-phone noise suppression as well as echo cancellation. It turns out that having two microphones helps to distinguish a speaker's voice from background sounds, if there's a chip that can analyze all the signals. The Audience A1010 Voice Processor, which is designed to handle that, is already available in one phone sold in Japan and one in South Korea, and it will soon appear in more handsets, according to the company.

To tap into the human ear's ability to distinguish between sounds, Audience had to first understand the ear as well as the human brain. It drew upon an advisory board of eight scientists, each of whom studies a different part of the brain. After identifying and diagramming the elements of human hearing, Audience designed a DSP (digital signal processor) with custom algorithms that carried them out.

The key to the company's chip is the Fast Cochlea Transform process, which attempts to do the same thing as the cochlea, a spiral chamber in the inner ear. The cochlea breaks up signals into different frequencies, allowing us to distinguish one pitch from another. Also, because we have two ears and two cochleas, the brain can figure out where a sound is coming from based on small differences in the loudness and timing of the sounds picked up by each cochlea, Watts said.

Here's how the two microphones work: When you're talking into a phone with two microphones, your voice goes mostly into the one nearest your mouth. The second microphone typically will be opposite the one you normally speak into, on the back side of the phone.

Noises from a long way away from the two microphones are about even in volume. Also, there's a delay -- maybe half a millisecond -- between when the background noises reach one microphone and the other. The A1010 factors in that delay and is also smart enough to figure out which incoming pitch belongs to your voice and which sounds come from other sources, Watts said, demonstrating the technology last week at Stanford.

"Once we've decided what is a foreground sound -- voice -- and what is a background sound, then we can suppress the background sound," Watts said.

Audience's chip goes beyond the technology used in noise-reducing products such as the headphones pioneered by Bose, Watts said. Those headphones identify steady sounds such as ventilation fans and then put out their own steady tones to counter them. The Audience DSP can suppress even intermittent and changing sounds, such as someone talking nearby you.

Suppressing background sound could be critical during, for example, calls to your spouse or boss to explain why you're late. A noisy bar could sound a lot like a quiet office. But once you've spent the extra money for a phone with two microphones and an extra chip, you also get a benefit yourself: The Audience chip can analyze the background sounds coming from around the person you're talking to and cut out that noise as well, Watts said.

All this could mean a big leg up for phone makers that latch on to this technology, according to Avi Greengart, who analyzes the mobile industry at Current Analysis. At the moment, vendors are all pushing their browsers and music players, he said.

"The one thing that hasn't been claimed yet is the mantle of, 'Our phones make better phone calls,' " Greengart said.

Audience's Watts said the chip could go into Bluetooth headsets, where some effort has already been made at cleaning up noises, such as in the sleek Jawbone from Aliph. But he pointed out that the cell phone market, about ten times the size of the headset industry, was more attractive.

The Audience Voice Processor is only 2.7mm by 3.5mm in size and sits between the microphones and the main baseband processor of a phone. It stands out from other audio chips in that it has to have both an analog-to-digital conversion piece (for bringing sounds from the microphones into the Audience chip) and a digital-to-analog converter to push the cleaned-up sounds to the main processor. Cellular baseband chips are used to getting sound directly from a handset's microphone as an analog signal.

Because it simply sends out an analog signal, the chip can be used with GSM (Global System for Mobile Communications), CDMA (Code-Division Multiple Access) or any other cellular technology, Watts said. It's currently available in the Sharp SH705ill handset in Japan and the LG Electronics LG-SH400 in South Korea.

In honor of the Audience chip's ability to conceal where secretive subscribers are calling from, SK Telecom is advertising the LG model as the "alibi phone."

To comment on this article and other PCWorld content, visit our Facebook page or our Twitter feed.
Related:
  
Shop Tech Products at Amazon