Why can't you just read the sound and find when the magnitude exceeds a threshold that you think is required for it to be audible. Like if your sound amplitude ranges from -1 to 1, and you think the value needs to be more than 0.05 to be audible, just do
[y, fs] = audioread(filename);
startingIndex = find(abs(y) > 0.05, 1, 'first');
