What is good, better, and the best audio? For WinHEC Cortana - - PowerPoint PPT Presentation

For WinHEC Cortana Workshop, Shenzhen, 11 August 2015

What is good, better, and the best audio?

2 Company Confidential and Proprietary

Laptop as a Speakerphone

 Enterprise Class Full Duplex Speakerphone Experience  Optimized for MS Lync/Skype  Flawless Cortana Accuracy  Audio Output >2X Louder  No Risk to Speakers  Better Media Consumption experience

mHD-Audio

• r

I2S / I2C

CONEXANT

CODEC

with Integrated AudioSmart Amp

CONEXANT AudioSmartTM Software

• SSP Input Noise Reduction
• Output Enhancement
• SoftAware Speaker Control

Original System AudioSmart System USB Speakerphone (benchmark)

65dB 81dB 76dB Sound Pressure Level at 1m

No other solution can deliver full duplex conference speakerphone experience while concurrently delivering this level of loudness from laptop or tablet form factor

2.0

3

Conexant Confidential and Proprietary

AUDIOSMART

TM

INPUT TECHNOLOGY

4

Conexant Confidential and Proprietary

Conexant Smart Source Pickup (SSP)

SSP: Conexant proprietary noise suppression algorithm. The one easy solution.

• Based on a Blind Source Separation technique, modified for real-world robustness.
• NOT Beamforming ; SSP uses spatial representation of target speech and noise

sources to reduce stationary and semi non-stationary interference (i.e., babble noise).

• Speech can be at any angle relative to the microphones.
• Noise can be at any angle relative to the microphones, even in the same direction as

the speech. Full-Sphere

“Noise” “User”

5

Conexant Confidential and Proprietary

SSP Voice + Speech Algorithm Value Proposition

Conexant’s Smart Source Pickup Input Pre-Processing Solution with 5M Technology Enables:

• High resolution VoIP + class-leading Automatic Speech Recognition

from up to 5 meters (16 feet) away!

• Clear voice quality calls and videoconferencing in real-world

conditions (i.e., across a large, noisy room).

Conexant Voice/Speech Input Pre-Processing

Speech + Interference Clear Speech

Increased VoIP Clarity

Decreased Word Error Rate Speech Recognition Engine

Intelligible Voice Calls

6

Company Confidential and Proprietary

Clamshell Stand Tent Tablet Lay-Flat

SSP Versatility

Supports all device orientations and use-cases with only 2 MICs, NO holes drilled in glass, and ZERO user input!

7

Conexant Confidential and Proprietary

AUDIOSMART

TM

OUTPUT TECHNOLOGY

8

Conexant Confidential and Proprietary

• Constantly delivers DSP-optimized signal to speakers at maximum safe

power level, independent of software

• Actual output power of H-bridge is monitored at chip pins (closed loop!)

and intelligently averaged 100% in hardware

• Closed-loop system actively monitors speaker power consumption
• Output RMS is compared to a programmable threshold matching exact speaker

capabilities

• RMS measurements drive intelligent gain control
• Adjustments are time-windowed to avoid false triggers by spikes in audio content

Class-D Power Optimization

RMS

PGA DAC

decrement gain during excess RMS output

Class-D Speaker Amplifier

9

Conexant Confidential and Proprietary

AUDIOSMART

TM

IOT EXAMPLE

10 Company Confidential and Proprietary

Speaker Performance Comparison

Amazon Echo

• Seven Microphones
• Amazon Voice Processing
• Amazon voice trigger engine

Conexant Smart Speaker

• Two Microphones
• Conexant Far-Field Voice Processing
• Sensory voice trigger engine

(embedded in Conexant SOC)

2.0

11 Company Confidential and Proprietary

Conexant Noise Suppression Performance vs Amazon

~40% Hit Rate improvement in noisy conditions (Noise 5dB lower than speech) No Hit Rate degradation in “clean” conditions (Noise 10dB lower than speech) 30% Hit Rate improvement in extremely noise conditions (Noise same level as speech) With Conexant processing, user will be understood one in three attempts With Amazon processing, user will almost never be understood

12

Conexant Confidential and Proprietary

AUDIOSMART

TM

ABOUT CONEXANT

13

Company Confidential and Proprietary

<2000 2006 2010 2012 2014

Acoustic Echo Cancellation enables speakerphone in production Fax/Modems Low-power PC audio codec in production computers DSP codec with integrated Class D in production systems Expanded PC audio portfolio (ADI acquisition) USB codec in production headsets, speakers and speakerphones DSP Codec for high quality Voice Input and Communication in Tier 1 TVs Windows APO Library complements HW portfolio

20+ years of Audio Signal Processing Expertise 20+ years of Audio Hardware Expertise

AudioSmartTM Class-D Integrated with Codec

History of Conexant Audio

14

Company Confidential and Proprietary

Summary

• Single supplier providing Silicon & Software optimized to work “better

together” – Silicon – High Performance ADC, DAC, CODEC’s, DSP & Smart AMP – Software – Input Processing, Playback Processing & Awareness – Ecosystems – Driving end user use cases with Industry leaders – Partnerships – partners telling OEM’s – “Conexant’s solutions work better” – Applications – support Audio for PC/Tablets/Headset/Docks/Smart Home/IoT

15

Company Confidential and Proprietary

Worldwide, World-Class Support

*Supported through a network of 20 Distributors and 13 Rep Firms

France

Headquarters Design Centers Sales Offices Design & Sales

India China Korea Japan Taiwan Singapore Waltham Headquarters Irvine Austin

• Highest concentration of

computer audio systems knowledge in the industry

• Expert Support AND

Engineering staff in every major time zone

• Proven track record enabling
• n-time shipment of Tier-1

OEM products