The research team recorded an entire week’s worth of class meetings of three courses at four times during the course of the semester. These class meetings were recorded in video from four different angles. This resulted in a very large amount of data (approximately 200 hours of raw video) requiring about 1 Terabyte (1,000 Gigabytes) of storage space. The audio portion of the video was also rendered out and compressed into MP3 format files for ease of use and analysis when research questions can be answered without video.
Cleaning Classroom Audio
The resulting MP3 files were much easier to process and store, but audio quality, which was lacking in the original raw video, still needed to be improved. In order to make the recording process unobtrusive, the cameras and microphones were installed at the ceiling in a large classroom. As a result, classroom audio recordings included a large amount of mid-frequency noise from building electrical and mechanical systems. This noise made it difficult to discern what individuals were saying on the recording.
In order to improve the quality of the classroom audio, I first equalized audio to remove reduce the volume of sounds at frequencies above and below those typical of speech. Frequencies below 1,000 Hz and above 7,000 Hz were reduced by 20 dB. Frequencies between 1,000 Hz and 7,000 Hz were reduced according to a generally bell-shaped curve centered near 2,800 Hz. Once audio had been equalized to focus on speech, the next priority was to reduce hissing, humming, and other constant noises which helped to mask speech. I used a noise removal tool integrated into the Audacity software package (v. 2.0.6). After selecting a noise profile for each track, I reduced noise by 48 dB on each recording (with a sensitivity of -1.00 dB and attack/decay of 0.20 seconds, utilizing a frequency smoothing window of 250 Hz). Finally, I converted each recording from a stereo recording to a mono recording and combined the recording for each camera – four total – into a single file, synchronized and balanced the audio sources, and exported the audio into a cleaned MP3 format audio file.
This effort resulted in a noticeable increase in intelligibility and reduction in recording artifacts, especially loud, high- and low-frequency pops and bangs which had limited the listening volume of the recordings.
Selecting Classroom Audio
The classroom audio used for this dissertation come from the first week of classes because this is the time period where the highest amount of socialization of new organizational members (the students) typically occurs.