There was a lot more said during the Apollo 11 missions besides "the Eagle has landed" and "that's one step for man." To help preserve and make accessible the thousands of hours of recorded mission audio, a team of researchers at the University of Texas at Dallas used speech recognition technology to unscramble and analyze the conversations between astronauts, mission control, and technicians across a quarter of a million miles of space.
Sometimes a technological advance can be a curse – especially if technology keeps galloping ahead at a breakneck pace. The analog audio tape technology available in the late 1960s must have seemed like a blessing at the time. The Apollo 11 mission was probably the biggest story since a curious fish decided to set foot on land 380 million years ago, but unlike that development, the first Moon landing could be recorded in more detail and more comprehensively than any other great event in history up to that time.
The curse was that it could be recorded in more detail and more comprehensively than any other great event in history up to that time. The sample used by the University of Texas project alone included audio from the Apollo 11, most of the Apollo 13, Apollo 1, and Gemini 8 missions and this partial collection comprised of 200 14-hour, 30-track analog tapes. That is a lot of audio to archive, never mind analyze and make available to future scholars.
Obsolescence was a particular problem because the tapes could only be played on a machine at the NASA Johnson Space Center in Houston called a SoundScriber, which used a manual crank to move from track to track. And on top of all of this, the recordings were raw. Nothing was edited or collated and there was nothing to show who was talking as many technicians spoke over one another.
According to Center for Robust Speech Systems (CRSS) founder and director Dr. John H.L. Hansen, using the Soundscriber would have required 170 years to handle the Apollo 11 mission tapes alone. Funded by a National Science Foundation grant, Hansen's team in cooperation with the University of Maryland developed a new sound system that could do the same job in months.
"We couldn't use [SoundScriber], so we had to design a new one," says Hansen. "We designed our own 30-track read head, and built a parallel solution to capture all 30 tracks at one time. This is the only solution that exists on the planet."
From left, Dr. John H.L. Hansen, Chengzhu Yu PhD'17, Dr. Abhijeet Sangwan and Lakshmish Kaushik pose with a model of an astronaut at the NASA Johnson Space Center in Houston
The next step was to spend five years transcribing and reconstructing the audio archive using new speech processing and language technology that could handle up to 35 people in different locations with many speaking in regional Texas accents. This required new algorithms that could recognize individual voices and analyze them to sort out who said what and when using a process called diarization.
The goal was not only to simply preserve, clean up, and transcribe the audio, but to learn more about how NASA personnel worked together under the stress of a manned lunar mission. This was particularly difficult because the recordings also needed to be placed in chronological order. But the team feels that this effort uncovers the story of the unsung heroes of the Space Age.
"When one thinks of Apollo, we gravitate to the enormous contributions of the astronauts who clearly deserve our praise and admiration," says Hansen. "However, the heroes behind the heroes represent the countless engineers, scientists and specialists who brought their STEM-based experience together collectively to ensure the success of the Apollo program," Hansen said. "Let's hope that students today continue to commit their experience in STEM fields to address tomorrow's challenges."
As part of the project, seven undergraduate senior design teams created the Explore Apollo website where the public can search the audio and transcriptions.
The research was published in IEEE/ACM Transactions on Audio, Speech, and Language Processing.
Source: University of Texas at Dallas
