Overview

Significant advances in automatic speech recognition (ASR) have led to an abundance of devices and applications that use speech as their main interface. Microphones are no longer regarded as mere voice input devices for phone calls but rather as acoustic sensors for human-machine interaction and for acoustic sensing of the environment. We have become to expect of our devices a high-quality ASR as well as other audio-related functionalities such as crystal clear voice calls and immersive high-definition audio.

As we are getting closer and closer to creating human-like machines, , we need to develop similar audio processing abilities to the ones we have as human beings.

  • Voice activation and speech recognition will become key in almost every device
  • Voice communication must be clear and high quality even in noisy surroundings
  • Audio processing allows the delivery of truly immersive sound
  • Sound sensing enables machines to interpret and act upon recognizable sounds

Capturing, processing and reproducing audio signals are fundamental to mobile devices, wireless speakers, smart home devices, automotive infotainment systems and consumer equipment.

Today’s smartphones, tablets and wearables require ultra-low-power, high-performance processing for speech recognition and reproduction, ‘always-on’ functionality and advanced multi-microphone voice capabilities. Smart wireless speakers and other smart battery-operated devices must be optimized for the best combination of performance and power consumption as they employ multi-microphone setups for far-field voice pickup, which can be hard on the battery if not implemented properly. And high-end audio for home entertainment and automotive infotainment systems demands the best possible hi-fi audio output.

The key to meeting these requirements lies in embedding multi-functional DSP IP architectures provide the following:

  • Provide for small die size, high code density and high processing power
  • Allow SoC designers to select the optimal implementation in terms of silicon area, power consumption and operating frequency
  • An ecosystem that includes development platforms to speed application development and prototyping as well as a wide variety of software modules available off-the-shelf

target markets

Mobile Voice

Reproducing a clean and noise-free voice signal involves using complex codecs such as EVS, as well as advanced noise reduction techniques.

Human-Machine-Interface

Human-machine interaction includes voice activation, speech recognition, and speaker identification/authentication. It often mandates ultra-low-power and ‘always-on’ solutions, some of which require far-field voice pickup, beamforming, acoustic echo cancellation, dereverberation etc.

Mobile Audio

Efficient audio processing ‘on the move’ is highly relevant these days not only for smartphones, tablets, and wearables but also for wired headphones and wireless headsets.

Smart Home

As devices for the home get smarter by the day, smart speakers, security cameras, thermostats, smoke alarms and many other home appliances introduce far-field voice control, high definition voice communication and high-quality audio.

Consumer Audio

Digital TVs, set top boxes, wireless speakers, game consoles and other consumer entertainment systems are expected to deliver immersive audio at ever lower price points.

Automotive

Automotive ‘infotainment’ systems integrate voice processing and speech recognition alongside high-end audio output in a challenging environment with a high noise floor.

CEVA's audio/voice/sensing DSP is at the foundation of our DBMD4 Always On Voice processor, delivering the desired power and processing efficiency to drive the industry's lowest power 'always-on' solution.
Ofer Elyakim, CEO of DSP Group