Research On Subjective And Objective Evaluation Of Virtual Bass Algorithm

Due to the physical size constraints of flat panel television loudspeakers,low frequency performance of these loudspeakers is generally limited and unsatisfactory.The equalizer method is often used to solve the problem,but simply increasing the low frequency energy through the equalizer causes signal nonlinear distortion,reducing the service life of the loudspeakers.In this thesis,the virtual bass algorithm is introduced to enhance the low frequency performance of the loudspeakers by inducing the human brain to perceive the fundamental frequency from its higher harmonics,solving the practical problem of poor low frequency playback capability of the flat panel television loudspeakers.The main work is shown as follows:Firstly,two algorithms of virtual bass are introduced:time domain algorithm-nonlinear device method(NLD)and frequency domain algorithm-phase vocoder algorithm(PV).The implementation of these two algorithms is carried out.Secondly,these virtual bass algorithms do not take into account the characteristics of the loudspeaker,which would lead to significant difference of subjective perception.In this thesis,two kinds of typical television loudspeakers are selected to evaluate the customized processing of the virtual bass algorithm by modifying the cut-off frequency of low-pass filter.The subjective experimental results show that each kind of television loudspeakers achieves the best perceptual performance when the cut-off frequency of low-pass filter matches the cut-off frequency of the loudspeaker.Finally,subjective listening tests of virtual bass sound quality are carried out by two subjective evaluation parameters of sound quality at low frequency:fullness and strength.However,the normal subjective listening tests are often time-consuming.In this thesis,an objective evaluation method is introduced for virtual bass algorithms based on the Model Output Variables(MOVs),the Objective Difference Grade(ODG),and the Distortion Index(DI)from the ITU Recommendation ITU-R BS.1387,as well as the Audio Spectrum Centroid(ASC),one of the low-level descriptors in MPEG-7 standard.The multiple linear regression analysis method is utilized to exploit these metrics.Compared with the objective evaluation method merely by ODG or ASC,the correlation coefficient between subjective and objective results has been significantly improved by this multi-parameter optimization hybrid model.The efficacy of the proposed method is validated by experiments.