DSP audio codecs focus on voice control and enhanced audio experience

September 19, 2014 // By Graham Prophet
Combining a low power voice trigger, multi-channel beamforming and accessory detection to produce breakthrough improvement in performance, Dialog’s DA7322 and DA7323 audio DSP CODECs are intended for applications with up to four analogue or digital microphones, respectively.

With a high level of audio digital signal processor (DSP) functionality, and with a minimum power consumption requirement, these products are specifically developed for voice-centric applications found in the latest generation of tablets, wearables, webcam accessories and in-vehicle entertainment.

The CODECs incorporate an integrated 24-bit DSP with a voice engine package that provides an improvement of over 35 dB signal-to-noise ratio (SNR) through the use of multi-channel beamforming, noise suppression and acoustic echo cancellation features. The result is an enhancement in speech performance with greatly improved sound quality for VoIP and voice recognition applications, which is particularly important for end users in a noisy environment.

Dialog says that independent of the ambient noise conditions the consumer is experiencing, both audio CODECs deliver a superior audio experience by reducing unwanted noise while maintaining voice quality during VOIP calls on applications such as Skype or Lync.

Targeting wearables, a low power voice trigger is available on the DA7322 and DA7323 where over a dozen keywords can be supported with over 95% accuracy without false triggering. This capability is demonstrated by the TrulyHandsfree always-on voice activation and recognition software provided by Sensory, which includes trigger to search, user-defined triggers, speaker verification and identification and multiple language support.

Beamforming technology built into the DA7322 and DA7323 voice engine package has been designed to offer maximum flexibility in terms of microphone placements in order to support both endfire and multi-MIC configurations. The beamforming algorithm also has the ability to suppress a variety of stationary and non-stationary noise sources.

The embedded DSP engine includes a suite of audio enhancement SmartEffects ranging from multi-band equalisation to stereo widening. It also contains a psychoacoustic bass boost that can be enabled to improve perceived low frequency response for playback on low cost speakers, while dynamic range control and volume levelling offers more perceived loudness and consistent audio levels.

The accessory detection integrated on the CODECs supports 3- or