AES Journal Forum

Thanks for the interest in the paper. I'll try to address all of your questions:

Section 4.1, Paragraph 2 & 3

- The selected modes were equalized using third-order Butterworth filters in a bandstop configuration. The cutoff frequencies are defined based on the measured spectral widths of each target room-mode

- After equalization the sound field exhibits less spatial variance between the two locations, but with reduced low-frequency content due to the applied equalization.

Section 4.1, Paragraph 4

- The test material was chosen for its rich bass content, which excites the room-modes in question. Of course the specific levels of modal excitation will vary among the signals.

- The algorithm operates consistently, regardless of program material, with the exception of the transient content detector which monitors the transient content of the input signal to dynamically weight the virtual bass algorithms running in parallel.

- The signals are roughly aligned by matching levels in the processed low-frequency band (before the PEQ is applied) to the levels in the unprocessed signal. Ideally, the virtual bass components should compensate for the absence of the removed components.

Section 4.1, Paragraph 5 & 6 + Section 4.2, Paragraph 1

- The subjects were free to move between the seats as many times and as often as necessary.

- Sound quality ratings were based on general impressions of the full-range sound field at each listening location. The scale ranges from 0 (poor) to 100 (excellent) with appropriate descriptors at intermediate intervals.

- The test subjects consisted predominantly of non-experienced listeners with no/minimal hearing loss.

- The subjects were free to choose three programs for testing. This method was chosen to allow the subjects to listen to material they are familiar with.

- The described listening test was designed to provide proof of concept, not absolute subjective performance. If the algorithm were picked up for use in a product a more rigorous subjective evaluation would be necessary.

- The data in Fig. 14 is the mean over the second and third quartile of the test data. As alluded to above, the bulk of time with this work was dedicated to algorithm development where the subjective tests were used for proof of concept. Again, more thorough testing and statistical analysis would be necessary if this were to be implemented beyond the current prototype.

Section 4.2, Paragraphs 2 & 3 + Discussion

- Since the most problematic room-modes (in terms of spatial variance between seats) have been removed, it is expected that seat-to-seat variance will be reduced after the virtual bass processing. The trick is to maintain acceptable sound quality after processing, which is largely dependent on the virtual bass essentially replacing the removed low-frequency components.

- Generally, the virtual bass processing can work in place of the physical low-frequency reproduction with the exception of the physical impact lost with removal of the components. This is what was observed during testing and is something that really can't be addressed with virtual bass alone (although this could be achieved using other means, but is beyond the scope of this work).