Low-Complexity Voice Activity Detector Using Periodicity and Energy Ratio

An alternative low-complexity method to identify voice and silence regions in a speech signal is introduced in this paper. Its performance, limitations, and some other voice classification techniques which deal with energy estimation are presented as well. The proposed algorithm uses periodicity measure, high-frequency to low-frequency signal energy ratio, and total voiceband signal energy to provide voice/silence classification. According to the experiments results the algorithm is less susceptible to variable acoustic environment. Furthermore, it is possible to use this voice detection scheme for very low-energy speech, which in turn makes the detection more robust in situations where a poor-quality microphone is used or where the microphone recording level is low.