V. Subba Ramaiah¹ and R. Rajeswara Rao^2
1Jawaharlal Nehru Technological University, Hydrebad and ¹Jawaharlal Nehru Technological University, Vijayanagaram, India

Multi-Speaker activity (MSA) detection helps in detecting the presence of whether the speech signals has a single speaker or multiple speaker speeches in the speech signal. It is easy to calculate the slope at ZCs (zero crossings) of the speech signal and makes a comparison with a suitable threshold (Th). Multi-speaker is declared as and when the zero crossing value exceeds the threshold. The impact of the proposed technique is compared to the existing technique by calculating the sample-by-sample ZCR (Zero crossingrate) value is demonstrated. Experimental results prove that the proposed ZCR technique achieves accurate results than the traditional techniques for MSA detection that uses the cepstrum resynthesis residual magnitude (CRRM) in the literature.

Feature Extraction, Multi-Speaker Activity Detection, Zero Crossing Rate