View Item 
        •   Utrecht University Student Theses Repository Home
        • UU Theses Repository
        • Theses
        • View Item
        •   Utrecht University Student Theses Repository Home
        • UU Theses Repository
        • Theses
        • View Item
        JavaScript is disabled for your browser. Some features of this site may not work without it.

        Browse

        All of UU Student Theses RepositoryBy Issue DateAuthorsTitlesSubjectsThis CollectionBy Issue DateAuthorsTitlesSubjects

        Speech recognition at higher-than-normal speech and noise levels

        Thumbnail
        View/Open
        ThesisFinal.pdf (2.096Mb)
        Publication date
        2010
        Author
        Gelder, M.E. van
        Metadata
        Show full item record
        Summary
        Previous research has demonstrated reduced speech recognition of normal hearing listeners when speech is presented at higher-than-normal levels (e.g., above conversational speech levels), particularly in the presence of speech-shaped background noise. The present study examined the interactive effects of signal-to-noise ratio (SNR), noise level and filtering on word recognition in noise. Speech recognition for 10 young adults with normal hearing was assessed for broadband, highpass and lowpass filtered speech (CF 1.6 kHz) at three noise levels (62, 72 and 82 dBA). The interfering noise conditions were steady talker spectrum matched noise, and 16 Hz interrupted noise with a duty cycle of 50%. The SNRs for the broadband conditions in steady noise were -12, -9, -6, -3 and 0 dB and SNRs 0, 3 and 6 dB were selected for the highpass and lowpass conditions. In interrupted noise, the SNRs for the broadband conditions were -30, -24, -18, -12 and -6 dB and for the highpass and lowpass conditions, the SNRs were -18, -9, 0 and 6 dB. For all conditions, speech recognition performance increased with increase in SNR, except for the lowpass filtered words presented in steady noise. At the same SNR, word recognition performance in interrupted noise differed significantly from performance in steady noise across noise levels. For broadband filtered words presented in steady noise, performance slightly decreased at higher SNRs (0, -3 dB) with increasing noise level. In interrupted noise, performance increased with increasing noise level for all SNRs, with significant interactions between SNR and level. Lowpass filtered speech was more difficult to recognize than highpass filtered speech and increasing the level of lowpass filtered speech at the same SNR did not improve speech recognition.
        URI
        https://studenttheses.uu.nl/handle/20.500.12932/4340
        Collections
        • Theses
        Utrecht university logo