The aim of the study was to determine an optimized single-number-quantity (SNQ) for airborne sound insulation, which predicts well subjective loudness of neighbor noise. Our study utilized the data of a recent laboratory experiment of Hongisto et al. 2014 (Acta Acustica united with Acustica). The experiment involved six different living sounds, wall consturction, and 59 subjects, thus forming a data basis solid enough for defining an optimized SNQ. The shape of the optimal reference spectrum was sought through optimization by formulating the problem as a nonlinear optimization problem with constraints, and solving it numerically. The new reference spectra were determined for six types of living sounds separately, and for living sounds in general. The optimized SNQs, which were based on the optimized reference spectra, associated much better with subjective loudness than the existing SNQs from ISO 717-1:1996 and ASTM 314 (1999). A suggestion for a reference spectrum for normal living sounds is given at the frequency range of 50-5000 Hz. Low frequencies are substantially less emphasized by the suggested spectrum compared with the reference spectrum of e.g. C_50-5000.