Post-processing speech recordings during MRI

We discuss post-processing of speech that has been recorded during Magnetic Resonance Imaging (MRI) of the vocal tract area. These speech recordings are contaminated by high levels of acoustic noise from the MRI scanner. Also, the frequency response of the sound signal path is not flat as a result of restrictions on recording instrumentation and arrangements due to MRI technology. The post-processing algorithm for noise reduction is based on adaptive spectral filtering, and it has been designed keeping in mind the requirements of subsequent formant extraction. Speech material was used for validation of the post-processing algorithm, consisting of samples of prolonged vowel productions during the MRI. The comparison data was recorded in anechoic chamber from the same test subject. Spectral envelopes and formants were computed for the post-processed speech and the comparison data. Artificially noise-contaminated vowel samples (with a known formant structure) were used for validation experiments to determine performance of the algorithm where using true data would be difficult. Resonances computed by an acoustic model and, similarly, those measured from 3D printed vocal tract physical models were used as comparison data as well. The properties of recording instrumentation or the postprocessing algorithm do not explain the observed frequency dependent discrepancy between formant data from experiments during the MRI and in anechoic chamber. It is shown that the discrepancy is statistically significant, in particular, where it is largest at 1 kHz and 2 kHz. There is evidence that the reflecting surfaces of the MRI head and neck coil change the speech acoustics which results in “exterior formants” at these frequencies. However, the role of test subject adaptation to noise and constrained space acoustics during an MRI examination cannot be ruled out.