Update to Thoughts on Interaction Modalities for Interactive Mirrors
Friday, 24 June, 2016 § Leave a comment
So in my previous post i outlined why voice commands are an essential component for a robust viable interactive mirror. In this post i want to highlight the complexities of voice recognition specially within natural user interfaces.
Natural User Interfaces require that the voice command be abstract and fluid; where the system should recognize natural speech in contextually dissimilar methods. This idea can be explained in detail through an example.
Example: User needs get today’s weather
A user may say the following to get today’s weather.
Whats the weather today?
How hot is it today?
What’s the temperature?
They all mean the same thing however they are said in completely ways. The system should be robust enough understand the natural speech and determine the question being asked.
This is where things get challenging. Traditional methods for speech recognition rely on key phrases that are recorded and trained into the system and the user is limited to his interaction based on these limited phrases. As such the user experience of products seems mechanical and doesn’t allow for any natural connection between product and the owner.
To combat this a system can be designed to use ANN (artificial neural networks) however this would require huge amounts of data for training the system not to mention the complexity of designing such a system in terms of both hardware and software.
As such i propose, voice recognition can be more easily accomplished by utilising existing natural voice recognition systems such as for Google Now, Microsoft Cortana, Amazon Alexa or Apple Siri.
Of the existing systems two appeal to me the most Alexa and Cortana. Alexa is part of the home automation system as such can be used to integrate into other devices. However since the system i intend to build on will run on some sort of windows or windows IoT OS cortana can be used to make the product self sufficient.
What to choose? What to choose? What to Choose?