Comparison of various technologies in the acquisition of speech samples for transcription purposes: listener perception vs. acoustic analysis of signal quality

The purpose of this study was to investigate the quality of recordings obtained with a dedicated recording device in comparison to more readily available devices. Professionals in the field of communication disorders need to identify and transcribe specific details about individuals’ speech and need high quality recordings for this purpose (Louko \u26 Edwards, 2001). The dedicated recording devices that provide high quality signals are not readily available and can be expensive. The available research on this topic focuses on analysis of voice characteristics such as fundamental frequency, amplitude, and stability of amplitude and frequency (Lin, Hornibrook, \u26 Ormond, 2012; Vogel \u26 Maruff, 2008). This study extended this research to acquisition of samples for phonetic transcription of speech sounds. This research addressed two questions: (1) What combination of microphones and recording devices provides the clearest speech sample in terms of acoustic analysis? (2) Do listener preferences align with the quality of the recordings based on the acoustic analysis? Participants in this study included 5 adults who provided speech samples for the study and 20 adults who served as listeners and judged the quality of the recorded samples. The participants included both males and females, ages 19-23. Participants had hearing within normal limits, were native speakers of English, and were free of any speech or language disorders at the time of the study. The acoustic analysis for this study yielded information about the signal-to-noise ratio (SNR) for each device and the amount of peak clipping across the samples. This research provides information about the quality of recorded speech obtained from a dedicated digital audio recorder, a laptop computer, and an iPad. The findings showed that peak clipping was not a factor in selecting a device because this occurred only one time across all of the samples. The results for the SNR showed the iPad technology combination had the highest SNRs but also the largest measurement variability. The Marantz technology combination had the lowest SNRs with the least amount of measurement variability. A different pattern emerged from the listener perception data. The listeners rated samples from all three of the devices approximately equally with regard to presence of noise in the signal and the signal clarity. Taken collectively, the results of this study suggest readily available technology, such as tablets and laptop computers, can be used to obtain high quality recordings of speech