Abstract: In a conference, considering the packets only from a set of selected clients can reduce the degradation of the quality of speech because mixing packets from all clients can lead to lack of clarity in the speech of any participants. The automatic selection should be smooth and should not cause frequent interruptions. A method of selecting the clients for mixing is suggested here based on a new quantifier of the voice activity called Loudness Number (LN). The dependence of the Loudness Number on the amplitude of the packet at present and the past activity is clearly brought out. The structure of the packet used has been explained. A method to avoid echo and enhance the quality of the conference is presented. The contributions of the paper are expected to aid in the implementation of H.323 recommendations for the Multipoint Processors (MP). A working prototype based on the proposed Loudness Number is already functional.