all | Alexa Lambda Linux Reference Design

Firstly, an AWS SimpleDB database was created to store the observations that the Alexa voice tagging skill generates. The database was created by the code shown below. This assumes that the AWS SDK has been installed on the machine. See AWS SDK for JavaScript in Node.js for how to do this. The Alexa skill's schema and utterance database includes three speech intents, OccupancyIsIntent, PredIsIntent, and TrainIsIntent. This code can be found here. When the user's speech triggers it, OccupancyIsIntent causes Alexa to return the occupancy prediction generated by the predict() thread as explained above, PredIsIntent causes Alexa to return the last true prediction, and TrainIsIntent associates a particular observation with a pattern using voice (i.e., voice tagging) and then stores it in SimpleDB. The Lambda Node.js code that services the Alexa speech intents includes three functions to support these intents. A function called trainInSession() handles the TrainIsIntent intent, the function called predInSession() handles the PredIsIntent intent, and the function called anyoneHomeInSession(), handles the OccupancyIsIntent intent. Currently, TrainIsIntent and trainInSession() uses 10 fixed patterns with the mapping to specific paths through the home as shown above. This forces the user to remember the mapping during training. Although acceptable for test purposes, a more flexible approach is required whereby the user is asked to provide the path name during training or is offered to select a path from a menu in case the path already exists. These enhancements will be added in a future revision of the skill. The thread msg_io() in the Raspberry Pi real-time software calculates the timestamped sensor data as described above and the Pi's server was returns that along with other information as JSON in response to a command from the Alexa skill running in AWS Lambda.