In the last decades, circular tranducer arrays are widely used in medical ultrasound diagnostics, and a variety of methods are currently available for the recovery of image obtained from the data collected from these arrays. It is silently assumed that the circular array is ``ideal'' in the sense of the shape of the circle, location of emitter/receivers, absence of delays in their activation, etc. However, due to the manufacturing reasons, this is always not the case, and these inaccuracies lead to somewhat incorrect interpretation of the collected data and hence, imprecise image reconstruction. To overcome such drawbacks, a pre-diagnostics procedure, referred to as calibration and aimed at the identification of the actual inaccuracies, is to be performed. At present, there exist a number of calibration techniques based on various methods of optimization. Keeping in mind the amount of emitters in the array (say, hundreds), most of the existing techniques are overly complicated, leading to very high dimensions of the corresponding optimization problems (hence, severe requirements to the computer memory) and very slow convergence times. In this talk we present several possible new approaches based on different optimization methods for the solution of the calibration problem and show that they seemingly lead to better results.