Formally, your main problem - existence of a systematic error at determination of velocity. Integration error grows linearly over time. This error can be removed if you know not only the initial, but also final length of a muscle.

The second problem is connected with use of the obtained data in model. Stiffness and viscosity of muscles depend on an innervation and are known too roughly. Because of it calculations for the specific patient the complex mathematical problem, which complexity  depends on a features of task .

not sure if this is what you need but velocity equals the time derivative of the length, v=d/dt(l), or, by means of integration

l = \int_{t_1}^{t_2}  v(t) dt .

[v: velocity, l: length, t=time, d/dt(.): time derivative]

I am not sure what kind of data you have right now. Maybe you can also check the Musculoskeletal model plus 3D motion capture to obtain the musculotendon length. We use Simulink to model this process in order to know the muscle fiber length from the already known musculotendon length.

 Thanks Dr Thomas for your response. In the muscle model I guess the initial length of the muscle fibre and stop guessing when the force between muscle fibre and tendon reaches equilibrium. In the next step, we can get the length of muscle fibre without guessing by numerically integrating the current velocity of contraction.  

This has been described in the link attached with this comment. But I am not quite sure how to implement this in Matlab.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1357215/

Hi Ye,

Thank you very much for your response. How do you calculate muscle fibre length from the musculotendon length?