6. Conclusions
In this article an innovative solution for streaming speech over Bluetooth Low Energy is presented. The so-called BlueVoice application demonstrates the feasibility of multimedia communication in constrained sensor devices acting as IoT nodes. In the article, a vendor specific Bluetooth LE profile for speech communication is defined, and the BlueVoice design is presented by considering communication roles of involved devices, audio processing and compression choices, packetization issues and bandwidth requirements. The BlueVoice application is composed of a Tx node and a Rx node, acting as Bluetooth LE Slave/Server and Master/Client respectively. Periodical notifications are sent from the Server to the Client after the connection is established. On the Tx node, MEMS microphone digital output in PDM format is acquired, converted to PCM and then compressed to ADPCM, resulting in a communication bandwidth of 64 kbps on the Bluetooth LE link. The pro- file also defines a side-information mechanism, requiring an additional bandwidth of 300 bps, for a total bandwidth of 64.3 kbps required by the application. In order to evaluate the performance of the proposed solution, the BlueVoice application has been implemented in real hardware devices. Particular focus has been devoted to the Tx node, which has been implemented as a fully digital system composed of a MEMS microphone, a microcontroller acting as host and a network module acting as Bluetooth LE controller. Performance evaluation reported in the article shows the feasibility of the solution for the IoT context in terms of power consumption, processing requirements and memory footprint. In particular, a power consumption of only 9.570 mW has been measured for the network module during audio streaming, while completely acceptable memory and processing requirements have been experienced. Moreover, the communication delay has been measured as a key performance indicator. Such a figure of merit results in an average of 15.80 ms and a maximum of 29.00 ms, showing that the BlueVoice application is able to reach very high speech quality values at the receiver side and therefore confirming it as a suitable solution for advanced applications in the IoT scenario.