HomeProductsProducts Details

Add voice recognition/speech-to-text feature to your SOC design

Date: 21/08/2013
Time is ripe for the keypad to say good-bye from your computers, mobile phones, consumer electronics systems. The natural interface such as touch-interface is already preferred over keyboard to interact with the computers. Spoken voice is even more a natural preference to interface with the electronic systems which includes car infotainment, mobile phones, computers, washing machines or any such machines/equipments, which human beings instruct. For the computer, consumer, wireless mobile devices the voice interface is becoming key differentiator for it to sell hot in the market. But for engineers, adding voice interface requires semiconductor ICs supporting that feature. To quickly integrate voice recognition feature in ICs, semiconductor chip designers looking for ready solutions rather than designing on their own. Handling natural interfaces requires higher and different processing power and enhancing hardware capabilities. This necessitates electronic product designers to have a processor with digital signal processing capabilities. SOC designers are looking for third party intellectual property in integrating DSP feature in their product designs.

We noticed three companies providing IP for the audio/voice DSP supporting voice recognizing speech-to-text algorithms.

1. CEVA is offering IP package called CEVA-HD-Audio supporting voice recognition. CEVA is partnering with Rubidium in delivering a speech processing solutions such as automated speech recognition, biometric speaker identification, and text-to-speech.
To know more visit:
http://www.ceva-dsp.com/CEVA-HD-Audio

2. HDClear from DSP Group process digitised audio/voice signals for speech recognition. HDClear uses 3-D processing to separate background noise from voice. HDClear feature another sensor along with two microphones for 3-D processing of voice. Other features include acoustic echo cancellation, user adjustable speech rate, and loudness equalisation.
DSP Group claims HDClear offers better accuracy in noisy environments such as for training and road compared to its competing technologies.
To know more visit:
http://www.dspg.com/English/Article.aspx?Item=891

3.The combined solution of voice processing software from sensory and HiFi Audio/Voice DSP (digital signal processor) IP core from Tensilica, a company owned by Cadence is available for voice trigger and recognition.

Tensilica claims its HiFi Audio/Voice DSPs are the industry’s most widely used licensable audio/voice DSP family, with support for over 100 proven audio/voice software packages. The HiFi Audio/Voice DSP is designed to run in ultra-low power mode to save battery life, waiting to detect sound based on Sensory’s Low Power Sound Detector.

To know more visit:
http://www.tensilica.com/products/audio