视频

Conexant AudioSmart™ Brings Noise Suppression and Echo Cancellation to Amazon Alexa Enabled Devices at CES 2017

Conexant AudioSmart™ Enables Voice Barge-In on Devices

Conexant’s AudioSmart™ technology can be used to enable a superior far-field, voice-enabled user experience by filtering out background noise and focusing in on what the user is saying, and can be used to enable voice barge-in on voice-enabled devices.

In this demo, we compare a factory-spec iPhone 6s and factory-spec Samsung Galaxy S6 to a modified Nexus 5 running Conexant AudioSmart™ software. The AudioSmart™ equipped Nexus 5 provides superior far-field speech recognition performance compared to the competing devices, and enables always-listening voice control functionality even when sound is emitting from the device’s speakers (barge-in).

Conexant & Sensory Partner For Industry Leading Barge-In Solution

Challenges:

  • Noisy Hands-Free Calls: Noisy driving situations, such as driving at high speeds and with open windows make hands-free voice calls impossible.
  • Inaccurate Voice Commands: A noisy environment can cause inaccuracy with voice commands causing a bigger distraction to drivers than it solves.

Benefits:

  • Clear Hands-Free Calls: Clear, intelligible voice, even in the presence of noise, enables natural, safe hands-free conversations.
  • Accurate Voice Commands: Robust voice command and control reduces driver distraction and improves the overall driving experience.

Conexant’s Solutions:
Full Duplex Acoustic Echo Cancellation: Part of Conexant’s VoiceSpeech Processors and AudioSmart Software solution, full duplex AEC suppresses the voice coming from the car speakers from going back into the microphones, eliminating echo for the far end person. This is done simultaneously while picking up the speech of the person in the car, allowing for a natural, two-way (full duplex) conversation.

Smart Source Pickup Noise Suppression: Part of Conexant’s VoiceSpeech Processors and AudioSmart Software solution, Smart Source Pickup (SSP) suppresses background noise from all directions to enable a clear voice signal being sent over the Hands-Free voice call or to the Automatic Speech Recognition (ASR) engine.

Clear Hands-Free Phone Calls for Automobiles

Challenges:

  • Noisy Hands-Free Calls: Noisy driving situations, such as driving at high speeds and with open windows make hands-free voice calls impossible.
  • Inaccurate Voice Commands: A noisy environment can cause inaccuracy with voice commands causing a bigger distraction to drivers than it solves.

Benefits:

  • Clear Hands-Free Calls: Clear, intelligible voice, even in the presence of noise, enables natural, safe hands-free conversations.
  • Accurate Voice Commands: Robust voice command and control reduces driver distraction and improves the overall driving experience.

Conexant’s Solutions:

Full Duplex Acoustic Echo Cancellation: Part of Conexant’s VoiceSpeech Processors and AudioSmart Software solution, full duplex AEC suppresses the voice coming from the car speakers from going back into the microphones, eliminating echo for the far end person. This is done simultaneously while picking up the speech of the person in the car, allowing for a natural, two-way (full duplex) conversation.

Smart Source Pickup Noise Suppression: Part of Conexant’s VoiceSpeech Processors and AudioSmart Software solution, Smart Source Pickup (SSP) suppresses background noise from all directions to enable a clear voice signal being sent over the Hands-Free voice call or to the Automatic Speech Recognition (ASR) engine.

AudioZoom Camera Recording for Notebooks/Tablets

Challenges:

  • Unintelligible Voice Recording Due to Background Noise: wind, traffic, crowd and machine noises all mask quality and clarity of recorded voice.

Benefits:

  • Clear, intelligible voice recording with suppression of environmental and machine noise
  • Voice capture focus works automatically with video camera direction and zoom
  • Robust technology allows flexible mic placement and geometry for optimal industrial design

Conexant’s Solutions:

  • Conexant AudioSmart Software Solution: Conexant’s AudioZoom solution operates similar to a camera’s video zoom, focusing the audio recording on the voice of interest, in the same way video zoom focuses on a face of interest

Enhanced Speakerphone Mode for Notebooks/Tablets/Phablets, CES 2015

Challenges:

  • Small, Quiet Speakers: Typical tablet/phablet speakers are too quiet to be heard from a distance in speakerphone mode
  • Large Amount of Echo: Person at far end of conversation hears their own voice or “echo” and gets worse as speaker is made louder
  • High Background Noise: Ratio of background noise to speech is high

Benefits:

  • Engaging listening experience with over double the loudness from small, quiet speakers
  • Clear voice quality even with loud speaker playback and background noise
  • Robust to device orientation and user position, even when device resting flat on table

Conexant’s Solutions:

  • AudioSmart Amplifier: Integrated into Conexant’s AudioSmart CODECs, AudioSmart Amp offers over double the loudness from the same small, thin notebook/tablet/phablet speakers
  • Full Duplex Acoustic Echo Cancellation: Full duplex AEC suppresses speaker output captured by the microphones, even when speakers are being driven at double the loudness with Conexant’s AudioSmart Amplifier
  • Smart Source Pickup Noise Suppression: Smart Source Pickup (SSP) suppresses background noise from all directions, even the same direction as the user’s voice, to enable clear voice quality, important when using speakerphone mode in noisy homes, offices, cafes or automobiles

Accurate Far-Field Voice Control for Set-Top Boxes, CES 2015

Challenges:

  • Loud TV Playback: Far-field voice control should work accurately while TV content is playing loudly
  • Unknown TV Processing: Though the media content is sourced from the STB, it undergoes unknown processing by the TV before being played out of the TV speakers
  • High Background Noise: In a home environment, background noise, e.g. air conditioner/furnace, is significant when compared to the user’s voice

Benefits:

  • Accurate speech recognition from a distance and from any direction
  • Suppresses loud background noise, even from the same direction as the user’s voice
  • Cancels TV playback, despite unknown TV processing after content played from STB

Conexant’s Solution:

  • Full Duplex Acoustic Echo Cancellation with Handling of Impaired Echo Reference: Part of Conexant’s VoiceSpeech Processors and AudioSmart Software solution, full duplex AEC suppresses TV speaker output content captured by the STB microphones, even when the content has undergone unknown processing (has been impaired) after it was output from the STB to the TV
  • Smart Source Pickup Noise Suppression: Part of Conexant’s VoiceSpeech Processors and AudioSmart Software solution, Smart Source Pickup (SSP) suppresses background noise from all directions to enable a clear voice signal being sent to the Automatic Speech Recognition (ASR) engine

Accurate Far-Field Voice Control for Smart TVs

Challenges:

  • Loud TV Playback: Far-field voice control should work accurately while TV content is playing loudly
  • High Background Noise: In a home environment, background noise, e.g. air conditioner/furnace, is significant when compared to the user’s voice

Benefits:

  • Accurate speech recognition from a distance and from any direction
  • Suppresses loud background noise, even from the same direction as the user’s voice
  • Cancels TV playback, despite being louder than the user’s voice

Conexant’s Solution:

  • Full Duplex Acoustic Echo Cancellation: Part of Conexant’s VoiceSpeech Processors and AudioSmart Software solution, full duplex AEC suppresses TV speaker output content while accurately picking up the user’s speech recognition commands
  • Smart Source Pickup Noise Suppression: Part of Conexant’s VoiceSpeech Processors and AudioSmart Software solution, Smart Source Pickup (SSP) suppresses background noise from all directions to enable a clear voice signal being sent to the Automatic Speech Recognition (ASR) engine

Keyboard Click and Screen Tap Noise Suppression

Challenges:

  • Call quality is affected by loud keyboard clicks or screen taps
  • Offending participant is usually distracted, thus typically does not mute

Benefits:

  • Eliminates annoying keyboard clicks and screen taps during conversations
  • Automatically mutes clicks/taps with no user intervention or mode change required

Conexant’s Solution:

  • Conexant AudioSmart Software Solution: Conexant’s advanced keyboard click/screen tap noise suppression significantly attenuates transient noise, during speech and silence, with little or no degradation of the speech quality. Traditional transient noise suppression heavily degrades the quality of the speech by removing un-voiced speech parts.

AudioSmart Amplifier with >2x Loudness From Small Speakers

Challenges:

  • Small form factor products with small speakers sound quiet with typical media content or in conversation

Benefits:

  • Louder speaker output improves user experience when listening to music, watching a movie or having a conversation
  • Achieves over double the loudness from small speakers compared to traditional Class-D amplifiers

Conexant’s Solution:

  • AudioSmart Amplifier: Integrated into Conexant’s AudioSmart CODECs, AudioSmart Amp offers over double the loudness with high fidelity from the same small, thin notebook/tablet/phablet speakers

Far-Field Echo Cancellation with Non-Linear Distortion

Challenges:

  • Main speaker is several meters far from the microphones
  • Echo is much louder than the speech
  • Poor quality speakers and high volume produces high non-linear distortion

Benefits:

  • Full duplex VoIP communication with excellent single talk and double talk performance
  • Accurate speech recognition while device is playing content from its speakers
  • Robust to non-linear distortion from low quality loudspeakers

Conexant’s Solution:

  • Conexant’s Full Duplex Acoustic Echo Cancellation (AEC): Part of Conexant’s VoiceSpeech Processors and AudioSmart Software solution

Far-Field Non-Stationary Noise Suppression in High Reverberation Environment

Challenges:

  • Main speaker is several meters from the microphones
  • The reverberation is very high (larger than 500ms)

Benefits:

  • Suppresses highly non-stationary noise, e.g. other speech
  • Robust in highly reverberant environments

Conexant’s Solution:

  • Conexant’s Smart Source Pickup (SSP) Noise Suppression: Part of Conexant’s VoiceSpeech Processors and AudioSmart Software solution