Nathan Eyring and Dr. Timothy Leishman, Department of Physics and Astronomy

Introduction

Over the past years research has been conducted to understand how sound radiates from musical instruments. This has been done by recording a sphere of 2664 data points around a player through the process of many repetitions. This sphere is called the instrument’s directivity This data can be input into room modeling software to predict how an instrument will sound in any given location even before it is built. However there is the legitimate question of whether or not there this new data will change how room model sounds, and if someone listening would care. This research was an attempt to answer this question. It had two parts, a program designed to help train listeners how to critically evaluate sound signals, and an evaluation of the room model.

Methodology

Listening Training Software – The listening training software was designed based off of research done by Sean Olive1. It has trainees match different equalizations of common songs with corresponding equalization curves. There are eight training sessions which each consist of four songs being tested over two varying sets of equalization curves. Trainees were required to finish this training over the course of three weeks, but were not allowed to perform two training sessions on the same day.

Room Model Evaluation – The room model evaluation has been developed, but it still has not been executed. It consists of three main parts: gathering of personal information; comparison between the live recording, room model with standard data, and room model using the new data gathered. Participants take this evaluation in an anechoic chamber using Sennheiser HD 650 headphones.

Results

Results of Listening Training Software – The listening training software recorded the success of each trainee. The most important score recorded is how well they did upon their first attempt at each training session. Not all participants completed the full training, but for those had completed many training sessions. The most important score was how many correctly answered on the trainee’s first attempt. There was range of possible scores of 0-4. Participants’ first sixteen scores were compared to participants’ last sixteen scores, and an average increase of 0.91 points was shown with a standard deviation of 0.09. This effect was more profound in those without previous audio experience, and those with previous audio experience had at times no improvement over training. The average score was 2.56 with a standard deviation of 0.17.

Current Results of Room Model Evaluation – The room model evaluation has not yet been completed, however the recordings have been made in the concert hall modeled. The data has also been input into the room model software. From an initial sampling of three people, all three agreed that the new directivity made a drastic improvement in the room modeling software, and that it better mimicked the real concert hall.

Discussion

Listening Training Software – The listening training software was shown to drastically help those who had no audio background. To improve an entire point on a 0- 4 point scale is a drastic improvement in the ability of trained listeners. This shows that the training was largely successful in creating listeners capable of being able to more accurately scrutinize audio sources. Most trainees finished at near the same skill level as shown by the standard deviation between final scores being only 0.17 points. However, this research showed that for more highly trained listeners more difficult training should be developed. Some did not increase their listening strength at all during this evaluation. For future research this is something that should be altered.

Room Model Evaluation – While complete data has not been taken of the room model evaluation, there have already been many interesting findings. Prior to the creation of the varying sound files it was thought there would be nearly no change between the new directivity and already existing directivity. However a quick survey found drastic differences between the new acquired directivity and traditional directivity used by the modeling software. This shows great promise for future research to be conducted.

Conclusion

While the full extent of the research still has not been completed, the research has already been able to show success in a beginning attempt to understand the effect of new directivity. The listening training was able to bring most listeners to a near equal skill level that was of much better than the general starting point. The room model was obviously affected and improved by new directivity data gathered. Further research will be conducted to better show how drastically this new directivity improves the room model. Nonetheless, it should not be overlooked that all signs already point towards the great value of the new directivities.