3RD INTERNATIONAL CONGRESS ON TECHNOLOGY - ENGINEERING & SCIENCE - Kuala Lumpur - Malaysia (2017-02-09)

Implementation Of A Cloud Processing Based Voice Communication And Noise Reduction Embedded System Network

The use of voice processing systems is widely known and it is a research field that has been explored over the last decades. With the current booming concepts such as Internet of Things and Cloud Processing, there are plenty of options for relying complex processing algorithms to be performed by remote processors. In this research, a new concept is introduced for a remote audio processing network with the use of a small embedded system network. The purpose of the implemented system is to save hardware costs at the expense of cloud processing services by having only low grade 8-bit microcontroller recording units on the end nodes connected to a more powerful small embedded system server that can perform signal processing locally, either with the aid of cloud based processing or the use of both as an optional processing function. The server can act as a local router that can manage different incoming nodes that might perform a previous handshake and device management through TCP/IP common socket protocols. While the connection of the end nodes is wired for implemented test, it is completely possible to turn the system wireless with the use of a WiFi. The end nodes encode the recorded audio to common 16-bit wave encoded file that is uploaded via the router. The router’s processing management and interaction is performed in a two stage cloud interaction: the first step consists of utilizing a file management cloud for sound file upload and download while a second script processing cloud interacts with the cloud file management system to retrieve the incoming files and perform noise reduction techniques for proper file download and output by the requesting node. In order to record and play the recorded sounds, the use of a simple microphone and speaker/earphones were used. The entire node to node communication did not exceed 2 minutes for a half-duplex communication. The intention of implementing this kind of system was to design a low cost inter-phone/broadcast system in a local area that could save the cost of a high end processor in the end node size by means of useful software libraries and tools that are currently available in the cloud. This system could also have the capability to send the recorded files to a cloud service in order to classify the input as a specific command in order to control remote devices by means of an Internet of Things system. If higher end processors such as the used for the router were used as end nodes, it could be possible to establish a full-duplex network with the aid of cloud noise smoothing functionality.
Bruno Senzio Savino, Mohammad Reza Alsharif, Yuuki Yabiku