Aim
The system must be able to intelligently adapt to various surround conditions and create a playlist of songs that are suitable and representative of those conditions. The playlist that is created from predictions must also not contain any duplicate songs, nobody likes to hear two songs played in succession.
How this is achieved
The Intel Edison first gathers data about a few surrounding factors in a room:
-
Amount of Motion
-
Intensity of Light
-
Number of People
-
Time of the Day
The data is then sent to the Machine Learning model on the AWS which predicts and returns a number of seed values which are then passed onto the Spotify API through the Intel Edison. A list of 10 songs are then recommended by Spotify and added into the ongoing playlist, any duplicates in the recommendations are deleted and the next closest alternative is being added to the playlist.
Applications
Automating music playback in cafes or pubs can be very useful to businesses as music choice can be a major factor in attracting the right amount of crowd. This also frees up manpower in choosing a suitable playlist for the business, moreover with the removal of duplicate songs in the prediction system, the song playlist combinations are endless.
DESCRIPTION
Aim
SYSTEM ARCHITECTURE
The figure above gives an overview of the system architecture that is being implemented in this project. The Intel Edison functions as a gateway in the system, gathering information and then relating the information to the AWS server. The AWS server which stores the Machine learning model gives a set of predicted values based on the training data set stored in the S3 database back to the Intel Edison. The Edison then does processing of the returned data and generates a prediced list of songs to be added to the current playing playlist through the Spotify API.
TECHNOLOGY
Light Dependent
Resistor
The amount of lighting in the room is monitored using a very simple light dependent resistor which is being fed into the analog input of the Intel Edison. The internal ADC converts the input value into an integer which ranges from 10 to approximately 800 for very high light intensity.
Number of
People
The number of people that is present in the room is done using two ultrasonic sensors that are placed at the entrance of the door to record the number of people entering and exiting the room. However for the demonstration, facial recognition software will run off the webcam to count the number of people in the room.
Motion
Detection
The motion detection algorithm has been adapted from the OpenCV package that is available as a python package. This is done using a UVC compatible webcam that can interface with the Linux OS on the Intel Edison. The camera used in this project is the Logitech HD920. In particular, the code that is running from the Intel Edison measures the amount of motion in the room over a given period of time and averages it out. This value which ranges from 10 to 2000 shows the amount of motion in the room, where the value of 10 represents no motion in the room at all.