Tech spotlight: Sonocent Clear Lecture™ audio enhancement suite
Roger Tucker is the founder of Sonocent and leads our programme of R&D activity.
Roger has a PhD in telecommunications and worked for many years as a senior researcher in the field of speech processing at Hewlett Packard before founding Sonocent. It’s fair to say, audio processing and audio quality are two subjects that he’s passionate about.
Here Roger talks about the features included with our powerful Sonocent Clear LectureTM audio enhancement suite, available now with Sonocent Audio Notetaker for Windows (coming soon for Mac).
Without further ado, here’s Roger.
The Sonocent Clear Lecture™ audio enhancement suite is the culmination of a number of years development of audio enhancement functions designed specially for the needs of audio note-taking in lectures.
Users of previous versions of Audio Notetaker will be familiar with earlier versions of some of the functions. Most of these have been enhanced for Clear Lecture™, though two (Click Reduction and Brighter) have been included in the suite unchanged from the way they worked in earlier versions.
The driving force for the enhancements has been the use of mobile devices for recording, which tend to be more variable in audio quality than the high-quality DVRs used in the past.
Unwanted noise removal
The first four functions - Noise Cancellation, Click Reduction, Low Frequency Cut, High Frequency Cut - reduce unwanted noises.
The Noise Cancellation system analyses the audio in between chunks to form a “model” of the background noise, which can be hiss, hum or the so called “cocktail noise” of a lot of people talking at once - anything with reasonably constant properties. This model is then used to reduce the background noise in the audio.
The degree of reduction is greater in our desktop app Audio Notetaker compared to our mobile apps. This is because it is on by default in the mobile apps and so needs to be more robust to inaccuracies in the noise model.
Click Reduction searches for clicks in the audio (caused by key or mouse clicks) and reduces them to the level of the surrounding speech.
Sometimes a very clear close-up recording will have sounds which trigger the click detector (like a nice clear “t” sound for instance), but this rarely happens on a distant recording.
Low Frequency Cut is designed to remove the rumbly sound you often get on mobile device recordings, or the thud noise you get when touching the screen. Speech does not contain these low frequencies, so it does not affect the actual voice. Consequently, we turn this on by default on both the desktop and mobile apps.
High Frequency Cut is useful for removing any annoying high frequency sound from any source - maybe a voice with too much “edge” or a high frequency background noise. It makes the voice sound slightly “rounded”, but does not affect its intelligibility.
The next two functions - Thinner and Brighter - change the timbre of the voice.
Thinner reduces the mid to lower voice frequencies and is useful when the voice is a bit boomy for any reason - e.g. when a portable PA system is in use. A byproduct of this is that it also removes low frequency sounds just as Low Frequency Cut does.
Brighter tries to restore the higher voice frequencies which may have been lost because of a poor microphone. It does this by boosting the high frequencies but at the same time applying just enough noise reduction to maintain the background noise at its original level.
Although last on the list, this is actually the most important function in the Clear Lecture™ suite when it comes to recordings made on mobile devices. Unlike DVRs, which boost the recording level when the voice is quiet, recordings on mobile devices can be very quiet if the speaker is at some distance.
Adjust Volume Automatically is actually a very sophisticated algorithm which judges how much to boost the volume based on both the level of the voice and the level of the background noise.
Its objective is to set the voice at a comfortable listening level whilst keeping the background noise in the background. Although this is easy if Noise Cancellation is on (since the background noise won’t be there), the algorithm is designed to work equally well with noise cancellation on or off.
It also adjusts to changes in the voice level. As soon as the voice level goes quiet for more than a couple of seconds, the volume starts to increase to take account of the new level. This is especially useful during questions. Similarly, if the voice level increases, the system quickly adjusts to maintain a comfortable volume level.
Try it for yourself
Sonocent Clear Lecture™ is included with Sonocent Audio Notetaker for Windows (coming soon for Mac). Download a free full-feature 30-day trial.
Work for an institution and interested in implementing Sonocent software? Shoot us an email: email@example.com
Sonocent Audio Notetaker free trial
Select the Windows or Mac icon, enter your email address and click ‘Go’