Audio quality is critical to achieving a fast and accurate transcription, with poor quality, noisy audio easily taking two to three times longer to transcribe than clear “perfect” recordings. In this blog we hope to highlight some of the challenges that TTP’s transcriptionists face, and the tools that help them to improve the chances of producing a high quality output.
Imagine, for example, that your audio files are interviews with machinery operators that are recorded on the workshop floor, with shopping centre patrons in a busy mall, or train commuters during a journey. These are situations – often received by TTP – that inherently introduce background noise (usually constant) into the recorded speech. While recordings with background noise can certainly be transcribed successfully, the noise is distracting to listeners, and forces the transcriber to rewind segments of the recording numerous times to ensure it was correctly transcribed.
One product that can improve the quality of an audio file is WavePad, from NCH Software (http://www.nch.com.au/wavepad/), and its most useful tool for this scenario is the “Noise Reduction” feature.
This feature allows us to take a “sample” of the background noise from the recording and apply a process known as “Spectral Subtraction”. Spectral Subtraction attempts to reprocess the file to reduce the impact of that specific background noise (thinking about the examples above, you could take a sample of the factory or train noise from part of the recording where nobody is speaking, and the algorithm will remove the majority of that noise while not adversely affecting the spoken content).
At TTP, we always encourage our clients to record in the most quiet and optimal environment possible. We also understand that it is sometimes very difficult to find an environment that allows for perfect recordings, and that the nature of business and research occasionally requires that recordings be made on location and in environments outside direct control. With tools such as WavePad in our toolkit, we have a better chance to transcribe those files effectively to produce transcriptions of the highest possible standard for our clients.
Unable to display Facebook posts.