Have you tried to set up a excitation table? Do you know how to do it? It won't do you much good if we just throw a circuit at you if you have no idea how it was designed.
Ok, now what kind of gates are you limited to use, can you use any kind of gate you want? I ask because the first thing i noticed was that the circuit posted is capable of attaining 8 distinctly different states, when only 4 states are needed which could conceivably lower the flip flop count to only 2. You might not be able to get away with that because of the timing requirements, but i though i would ask anyway.
Im able to use any gate but must use as few as possible. With my first attempt i used only one AND gate on the k2 and QNot1 on the 1st and 2nd bistable I came to this through the k-maps that I worked out on 3,4,6,7 count sequence but the count still started at the original 7 value.