This paper investigates the problem of output feedback control for a class of stochastic feedforward systems with unknown control coefficients and unknown output function. Since many unknowns occur into the systems, all the states of the systems are unmeasurable or unknown which means that they are not available in the control scheme. To compensate these unmeasurable/unknown states, a new form of K-filters with time-varying low-gain is introduced in our design. As long as output function belongs to any close sector included in the maximal sector region of systems' output function, we can design a time-varying output feedback control law by integrating the well-known backstepping framework with the time-varying technique. Later, based on the improved LaSalle-type theorem, we analyze the regulation of the closed-loop systems. Finally, an induction heater circuit system with unknown inductance capacitance is given to show the effectiveness of our control scheme. (C) 2017 Elsevier Ltd. All rights reserved.