Midterm Proposal - ltsimps/Midterm GitHub Wiki

                                     Acme

Metis

“Words are, in my not-so-humble opinion, our most inexhaustible source of magic. Capable of both inflicting injury, and remedying it” Albus Dumbledore.

Here at Acme we have striven to integrate robotics and human interaction to improve the lives of our users. Our new product Metis is a product that listens to our users, how they talk, and shows them how they feel. Imagine a robot that is able to tell you how lively a conversation at a party is by turning orange. Imagine a visual indicator on how you and your significant other speak to other, perhaps encouraging better behavior if words are used to harm. Metis is a sensing robot that uses microphones to record conversations. With the words it ingest’s it will analyze and determine sentiment and show a visual indicator on what has been said gradually starting from a neutral position and changing on positive or negative scales of color. Metis is also a database because stores words, so the user can be shown which words have been used most frequently.

This isn’t a robotic mood ring, we see Metis as a voice platform. With the ability to record sound and distinguish conversations, Metis can tell an advertiser when someone needs a pick me up, or when someone cheers or laughs during a Television program. This product could also integrate with other platforms like Alexa, or or other IOT devices.

Development:

For the development of this product the SIP methodology will be used with a test driven approach. This product will be modularized in four components which are hardware, analysis, input, and output. i will focus on the perception piece of the Metis robotic system. The Metis platform can be extended to show Human like facial expressions as well as coloring and mood. The Metis Robot can function as counselor if words are extremely negative then it can interject by buzzing or speaking by to the user. If human intervention is needed, Metis can be operated remotely by a human.

Input: This component will handle all raw audio input and process that raw audio into a format that’s suitable for analysis.

Analysis: This component will handle all analysis on processed input and give an estimate out perceived emotion from audio input.

Output: This will handle all interfaces that will connect to the output and show the perceived emotion in textual and visual modalities.

Hardware: Metis will interface with Arduino or Raspberry PI changing colors based on the output from the analysis from Metis.

All components will be unit tested to mitigate risk in development.

Algorithms and Libraries: Frequency analysis of words will be used to indicate emotional content of audio input. Sentiment libraries will also be explored for their usefulness. For example bit text shall initially be explored for API use. This API is closed source.

Speech to texts libraries will be used to convert Audio input to textual input, for sentiment analysis. The Google Voice API will be explored.

The risks with both API, is integration with C++ and ROS indigo.

Deliverables: Arduino Robot that changes color based on audio input.

Risk:

Timespan of developing hardware and software component. This will be mitigated by developing code that will output textual indicators of audio input if there isn’t time for hardware integration.

Integration of third party libraries. This will be mitigated with early best effort prototypes with simple examples to see if the libraries can be used in a cohesive manner.