Word error rate over ASR-based speech quality measure (ASQM) values. Each data point corresponds to a specific noise type, SNR, beamforming angle, and environment (office or anechoic). Color/symbol denotes the SNR. The solid curve in each panel denotes a second-order polynomial fit to the data points between the according ASQM and word error rate. The correlation value r between fit and data and the goodness of the fit in terms of root mean square error E are given in the left bottom corner of each panel.