Real-time digital sound spectrum and RT time estimation

Based on musical acoustics, psychoacoustics, architectural acoustics, room acoustics and digital audio technology, Using geometrics and psychoacoustics, an underline sound remodeling method was proposed by compressor, amplifier, exciter and equalizer, etc. Frequency spectrum analyzing was realized by FFT discrete transform and Hamming window function. Compressor ratio was calculated by advanced algorithm in large scaled sound clips. Equivalent absorption coefficient algorithm optimized complications of material absorption coefficients. Integral weights were calculated in 24 octave band of audible domain which contributed to the balanced algorithm reported for the first time. RT time (reverberation time) was computed by the equivalent absorption coefficient with advanced Sabine function. Frequency spectrum and formant data samples were summarized according to different state variables, RT time and compressor ratio. A BP network was realized by several equalizer parameters adjusting. 20 groups of top 3 relative formants (60 data) corresponding to different thrusts of motor engines were proposed for the net training and the output layer of the BP network was realized by 3, 4 and 6 nerval cells. Network with 4 nerval cells was analyzed to be effective in both runtime spectrum curve matching and computing efficiency.