Rapidly evolving technology continues to challenge how we, as consumers, interact with our devices. As the touch technology matures, thanks to great improvement in capacitive touch in recent years, user interface designers are turning to voice recognition as the next big break though in how we interact with our most used technology. From “Okay, Google” to “Alexa”, the voice control and recognition is starting to be a key feature, but simply adding voice is not enough for a great user experience. Voice recognition must be robust, fast to implement and low power to enable always-on listening.
Making voice recognition robust requires the highest quality data work on. Voice data is best served by a strong DSP to clean up the audio before it is processed by the voice recognition engine. Voice data’s signal-to-noise ratio can be dramatically improved by leveraging processing such as beam-forming, noise reduction and echo cancellation. Processing like this involves a lot of mathematical computation best served by a DSP, such as the ultra-low-power C55x family of DSP.
As voice recognition is being added to many existing systems, it’s critical that enabling voice recognition and associated pre-processing is fast and easy. As rapid prototyping is a priority for many developers, we have created a TI Design for always-on speech recognition on our newest TMS320C5545 ultra-low-power DSP. Offering easy integration into any system with a tiny package size of 7 mm x 7mm, adding voice recognition to any system is only a download away!
Learn more about designing with TMS320C5545 DSP: