|
 |
dspZONE Products for the week of November 17, 2008
Audience Says…
Raising the Bar for Advanced Noise Suppression, Enabling Mobile Phone Users To Hear And Be Heard In Noisy Environments
Company begins sampling new Voice Processors to mobile handset manufacturers
Audience raises the bar for advanced noise suppression with its new voice processor that includes a full suite of industry leading voice quality enhancements that enable mobile phone users to hear and be heard in noisy environments. The Audience A1024 Voice Processor is now sampling and provides industry-leading transmit and receive noise suppression, acoustic echo cancellation, and voice equalization.
To hear the superior voice quality of the new A1024, go to http://www.audience.com/demo-transNoiseSuppressOff-pr.html.
According to a vice president in the LG Electronics R&D Center, “Our customers are delighted with the voice quality of LG mobile phones that include Audience’s noise suppression technology. The Audience voice processor provides the most advanced noise suppression available in the market, and we will continue to partner with them to offer our customers superior communications experiences in the noisiest places.”
New Voice Processor Offers Comprehensive Feature Set
While other industry vendors are working to improve their transmit, close-talk noise suppression, Audience now provides not only the most advanced noise suppression available but also an entire suite of crucial voice quality features including:
- Unprecedented instantaneous transmit noise suppression that provides consistent and reliable noise suppression of up to 30 dB (enough suppression to remove the noise of coffee grinders, blenders, background music and conversations from a busy coffee shop)
- For the first time in the market, up to 18 dB of receive non-stationary as well as stationary noise suppression from the far end - allowing mobile phone users to receive calls from noisy environments and hear the other person better
- Significant improvement in speakerphone capability with Audience’s superior Acoustic Echo Cancellation (AEC), lets mobile phone users have convenient hands-free voice conversations or video telephony calls without double-talk or echo even in noisy environments. The Full Duplex Type 1 AEC (per ITU-T P.340 specifications) provides industry leading Echo Return Loss Enhancement (ERLE)
- Voice Equalization that boosts and equalizes the incoming voice based on the level and for the duration of local background noise as the user moves through different noisy environments like a train station or a busy street
“Voice quality is fundamental to satisfied users of mobile phones,” said Will Strauss, president and principal analyst at Forward Concepts. “And, they want good voice quality no matter where they are. Cell phones with Audience’s voice processor give consumers the noise suppression features they need to hear and be heard in any location despite the noise level.”
With the A1024, Audience continues to make strides in ease of integration with small footprint, industry standard interfaces and support for analog or digital microphones. Further; with its new FlexMic capability and Auto-Calibration capability, Audience gives handset manufacturers more control over microphone placement during the design phase and eliminates costly microphone calibration during the production phase.
- Packaged in a 40 pin WLCSP with 0.5 mm ball pitch, the Audience A1024 Voice Processor connects directly to any baseband processor via standard interfaces and occupies less than 25 square mm of board space. Multiple baseband reference designs are available to support customer design and development activities
- Flexible front/back dual microphone configuration provides maximum flexibility to handset manufacturers for ID design and performance. Unlike beam forming solutions that require specific front/back mic locations, the A1024 supports a wide range of microphone placement, making it versatile for all popular handset types (candy bar, slider and clamshell)
- Audience’s Auto-Calibration feature takes input from the two microphones, and uses patent-pending techniques to equalize the signals, eliminating the performance variance due to microphone sensitivity. There are no adjustments required on the production line, saving significant production time and money
Audience Voice Processors are the first custom ICs that are modeled after the most efficient and accurate auditory system, the human hearing system. By understanding the auditory pathway – from the cochlea to the brainstem to the thalamus and cortex – Audience is the first company to deliver a commercial product based on the science of Auditory Scene Analysis (ASA), or the grouping and processing of complex mixtures of sound. Because the Audience Voice Processor handles signals the way people actually perceive specific sounds, it is able to identify and suppress noise sources in an extremely efficient and accurate manner.
“With our first voice processor, the Audience A1010, we established ourselves as the leader in transmit noise suppression performance, allowing mobile phone users to place calls from noisy environments,” said Jennifer Stagnaro, vice president of marketing at Audience. “The A1024 raises the bar again, delivering unprecedented transmit noise suppression of up to three times more suppression than the A1010 as well as other significant voice quality features.”
“In fact, the level of suppression is so high, industry standard tools cannot correctly measure it,” Stagnaro added. “Audience is working with industry standards bodies, including the CTIA Working Group, to modify the methods to measure higher levels of noise suppression more accurately.”
EN-Genius Says…
Audience’s voice processor is a nice example of how embedded signal processing can be used to improve the performance of an existing system with relatively little cost or power penalty. If the web-based audio demo of their new A1024 (link above) is anything like the device’s real-world performance, the processor is a big improvement over the noise suppression systems being used in today’s phones and could quickly become a must-have feature in premium (and maybe not-so-premium) products. Besides the obvious benefits to the people you’re talking to, the improved speech clarity it delivers could greatly improve the success rate of the current generation of voice recognition dialers.
The engineers at Audience explained to me that they spent a fair bit of time reverse-engineering the human auditory system to develop the design elements that went into the A1024. They explained that the brain uses the cochlea as an active filter to extract a constellation of auditory cues such as spatial phase relationships, pitch, and onset time which can be used to determine whether a particular signal element is of interest or should be filtered out. Audience created an electronic analog that mimics the phenomenon by adding a second mike to collect ambient sounds for a different location. It then uses an FFT-based transform that includes physical characteristics of the cochlea to identify signal patterns that most likely contain speech and eliminate unwanted background elements. According to Audience, the algorithm is powerful enough that it can extract and clean up voice signals even when the background noise is at an equal magnitude.
The noise suppression algorithm runs on Audience’s custom-designed DSP that has an instruction set which was developed specifically to handle their specialized cochlear transforms. Its pipeline architecture delivers one operation per clock cycle and uses a hardware accelerator core to handle the compute-intensive parts of the transforms In a cell phone, the A1024 is inserted into the signal path between the phone’s microphone and the main DSP using the existing PCM interface that connects the baseband and voice processors. Typical power consumption it 25 - 30 mA, depending on what functions are being used at the time. Audience says that, in some applications, the processor can actually reduce overall system power consumption in noisy environments because it eliminates unnecessary transmitter activity by reducing false voice activity detection.
Audience has packed enough extra processing power in its DSP that it can support several other important functions including some basic noise suppression for received voice signals and up to 1000 ms worth of full-duplex echo cancellation for speakerphone applications. While all handsets with speakerphone capabilities already have echo cancellation of one sort or another, it’s usually handled by the baseband processor and there are usually not enough spare processor cycles to support full-duplex operation. Audience also offers an optional voice equalization feature which senses the background noise level at the receiver’s ear and automatically boosts the incoming volume level to match. This could be a very handy feature for mobile phone customers who routinely use their phones in crowded office environments, busy streets, or industrial settings.
Audience says that the A1024 voice processor is more adaptable than its predecessor and has far fewer restrictions on microphone placement or sensitivity variations. Variations in signal strength and phase information due to microphone placement are accommodated by developing a calibration table for a particular case/placement profile.
While Audience is focused primarily on getting its processors used in handsets, they do admit that its noise canceling capabilities make it an ideal candidate for VoIP appliances and cellular car kits. The A1024 might even make sense for Bluetooth headsets if the 25 - 30 mA that it draws fits within your power budget. In any of these applications, the improved levels of voice recognition that the A1024 noise reduction makes possible could be the killer application that is worth the price of the chip alone.
The A1024 Voice Processor is sampling now. The voice processor can be purchased through Audience directly, or through the company's international representatives and distributors. Sample pricing ranges from $5 - $7. Audience was guarded about its pricing for high-volume production but I’d estimate that, in 100 k+ quantities, you’ll see pricing in the $2.50 - $3.50 range. For cost sensitive handsets, the A1022 Voice Processor with an optimized feature set is also sampling.
Product Page
|
|
|
|
|
