AUDIO DESIGN FOR UNITY - COMPARISON OF DIFFERENT AUDIO SPATIALIZERS
When working in immersive audio formats, a high-quality audio spatializer is needed to create an immersive and realistic experience. Built-in, 3D sound engines often do not provide satisfactory results. Therefore, the question of which spatializer-plugin to choose frequently occurs at the beginning of a new project. This is not a trivial question. There are several options available but it is hard to find comparative studies on this topic. For this reason, I want to share with you a master thesis by Veli Laamanen that I recently found while researching content on this topic.
CONDUCTING AN EMPIRICAL ONLINE TEST TO FIND THE BEST SPATIALIZER
While working on his project, Veli also came across the problem mentioned above. To find the best 3D audio spatializer, he conducted an empirical online test in which he asked the following question: “In terms of perceived directional precision, what is the most highly rated audio spatializer-plugin, and how much can its usage increase immersion when compared to non-spatialized audio?”
He started his research by searching for alternatives to the default built-in Unity plugin by Oculus. Other determining factors included: a focus on direct and easy availability, proper customer support and frequent updates. Finally, he put together a list of six different companies to compare their products within his empirical investigation.
His list consisted of the following companies and products:
- Oculus (Unity Default)
- Resonance (Google)
- Steam audio (Valve Corporation)
- Realspace 3D audio (Visisonics)
- VR Audio Kit (3d Sound Labs)
To gather as much data as possible, Veli chose an online test in which he asked the participants to listen to each different audio spatializer plugin using their headphones as well as to evaluate the perceived directional precision. He divided the test into two experiments in which every participant used the same audio material.
In the first experiment, Veli created an audio scene in which a moving virtual sound source circled the listener. After listening to the audio scene, he asked the participant to rate the directional quality of sounds being emitted from left/right, front/back, above/below, as well as the overall sonic quality of the audio scene.
In the second experiment, the “direction test,” all the participants were asked to choose the perceived location of the audio information within the 3D space. They were able to choose from six different directions in order to properly complete the “direction test”. The test played the 3D sound locations in random order, and a multiple-choice grid gathered the answers.
Since all the examined plugins differ from each other in their respective features, the test focused on the spatializing aspect only. Room simulations, reverbs and different tone colorations like equalization were disabled during the experiments. Fortunately, Veli collected a total of 98 answers of which 78 were usable in the end.
THE RESULTS OF THE "QUALITY TEST"
Now we come to the really exciting part of Veli’s research. First, I want to show you the results of the initial test, the “quality test”. The first figure shows an average Likert ratio, calculated from the mean values derived from the four questions. The corresponding scale is displayed on the x-axis and goes from 1 (poor) to 7 (excellent).
Next, the total average of the subject ratio for the quality test is displayed on the same Likert scale.
THE RESULTS OF THE "DIRECTION TEST"
Finally, you can see the results of the “direction test” below. For the evaluation, Veli calculated the percentage of correct answers per direction for each tested plugin. The graph below displays this percentage on the x-axis.
HOW IS DEARVR UNITY DOING SO FAR?
As we can see from both experiments, dearVR UNITY was the most highly rated spatializer of all examined plugins when it comes to the quality of the perceived direction.
THE FINAL "IMMERSION" TEST
Veli did not stop here. He conducted further research to find out how much a spatializer plugin influences the perceived immersion within a VR project compared to non-spatialized audio. For this immersion test, Veli used dearVR UNITY since it was the highest-rated plugin in the previous test. For the test platform, he used a prototype of his virtual heritage project: A virtualization inspired by Sammallahdenmäki, a Bronze age burial site that the UNESCO World Heritage list inscribed in December 1999.
To ensure a meaningful testing environment he made sure that all interactive audio sources were played back without room and reverb simulations as well as with static sound levels. On top of that, visualization of the whole scene was removed so as to limit the participants’ perception and thus ensuring focus on the audio signals.
Veli then presented two different versions of the virtual heritage project to the attendees, one version utilizing the dearVR UNITY spatializer and the other Unity’s proprietary amplitude panning methods. According to Veli, the central question behind the immersion test was: “If your level of presence in the real world is “100” and your level of presence is “1” if you have no presence, how would you rate your level of presence within tested virtual environments?”
Since measuring immersion of audio poses the challenge of the ambiguous definition of the word “immersion” itself, Veli held discussions with all participants to ensure that everyone understood the meaning of immersion (a simulative and measurable part of your level of presence that includes spatialized audio).
The table below shows the results of Velis’ investigations. Seven out of the eight participants felt that a virtual environment with spatialized audio was more immersive than the environment with the use of amplitude panning methods. The version with spatial audio rendering was rated 15-60 percentage points higher when related to the “level of spatial presence”.
In his conclusion, he stated that the usage of the dearVR spatializer plugin has a significant positive effect on the experienced immersion when compared to an identical virtual environment with non-spatialized audio:
“The improvement in the overall quality of 3D audio, when switching from Unity’s default spatializer Oculus to dearVR, was significant. I considered dearVR to be the best spatializer in terms of precise audio spatialization, from the first time that I tried all the spatializers.”
At this point I would like to encourage you to take a look at the thesis yourself since it contains a lot of other fascinating information about audio production for VR, basics of virtual acoustics and problems that arise when dealing with 3D audio.