The information about hypocenters is essential for engineering tasks to predict a danger created by earthquakes. A nonlinearity in the earthquake location problem, a one-sided distribution of seismic receivers lead to a search of solutions improving hypocenter location accuracy. The original earthquake location method has been developed by using a physical and geometrical representation of the time difference between first signals from seismic waves registered by three different receivers. Data processing is performed applying algorithms of a computational geometry. The solution error is formulated in terms of an analytical geometry. Thus the method differs from traditional approaches, which are based on a standard statistic error. Taking into account the promising testing results we guess that the geometrical method can be as an additional computational technique or as a tool to get the trial hypocenter, which is an input parameter of standard software for hypocenter determining.