We developed an integrated technique for estimating crustal anisotropy with a horizontal axis using radial (R) and transverse (T) receiver functions. The technique includes computing three individual and one joint objective function (JOF) and a reliability analysis of the estimated anisotropy. The individual objective functions (IOFs) are designed to: (1) maximize the peak energy of the stacked R receiver function after a cosine moveout correction in the Ps arrival time; (2) to maximize the correlation of the radial receiver functions after a full correction of anisotropy or (3) to minimize the total energy of transverse receiver functions stacked after a removal of crustal anisotropy. The JOF was computed by a weighted average of the three IOFs, and the reliability analysis uses the principle that stacking coherent signals can lead to an increase of signal-to-noise ratio. We applied the technique to synthetic receiver functions generated with 30–60 per cent white noise from a variety of anisotropic and heterogeneous models. The synthetic tests indicate that the proposed technique has good capability to recover the input models. Despite the presence of random and other coherent noises, such as those caused by inhomogeneous structures, in the data, the technique can always provide accurate estimates of crustal anisotropy. We applied the technique to two permanent seismic stations in western China and found significant crustal anisotropy beneath one station located at the northern edge of the Tibetan plateau. The observed fast polarization direction at this station follows the direction of the maximum horizontal tensile stress, suggesting that the observed seismic anisotropy is likely caused by mineral alignment in the lower crust. The station situated in the Sichuan basin, on the other hand, shows little to no seismic anisotropy, which may suggest that the crust beneath the basin is nearly rigid with very little deformation. The developed technique can be applied to any broadband seismic stations that have a good backazimuthal coverage of teleseismic events.