Detection of Robocall and Spam Calls using Acoustic Features of Incoming Voicemails

Proc. Mtgs. Acoust |

Published by POMA | Organized by ASA

PDF

Abstract: Spam communications are organized attempts mainly aimed at marketing, at spreading false information or deceiving the end recipient. Prerecorded messages called Robocalls are increasing every year, striking US residents with 46 billion calls in 2020. Carrier and telecommunication regulators have automated systems in place, that identify unwanted calls using Call Detail Records, which include call origin or call duration information, but the actual audio content is often overlooked. We propose an audio-based spam call detection method that uses acoustic features of recorded voicemails to identify Human calls from Robocalls, and to identify Spam calls from Non-Spam calls for human callers. Results show that voiced and unvoiced audio content carry sufficient discriminatory information to distinguish between Human and Robocall, but also between Spam and Non-Spam calls. Distinguishing between Human calls and Robocalls achieved 93% accuracy, compared to Spam vs Non-Spam achieving 83% accuracy. We expect that our automated approach can serve as an auxiliary tool, in combination with other call behavior statistics, to reduce the frequency of unwanted calls or fraudulent incidents.

 

Table 2 robocalls spam calls stats

 

Table 3 robocall and spam call classification