For a relatively easy answer you can use a short-time Fourier transform and use the largest value as the estimate of your sinusoid frequency. The basic idea is that you do short Fourier transforms at different points in time to see how it changes. The problem with that is that you have a tradeoff between time resolution and frequency resolution. Shortening the Fourier transform improves the time resolution at the expense of the frequency resolution. Conversely, making the Fourier transform longer improves the frequency resolution at the expense of the time resolution.
The frequency resolution can be improved by using quadratic interpolation to get a finer estimate of the true sinusoid peak.
Another way to go that can get you very good frequency estimation at every point in time is with a phase-locked loop. Phase-locked loops are a little trickier to set up though, because you have to get the feedback gains right. There is also a tradeoff in regards to responsiveness (akin to time resolution) and resistance to noise.