Audio coding schemes such as MP3 or MPEG AAC can use a variable number of bits to encode each signal frame. However they can still be used oil constant bit rate channels thanks to a variation buffer limited in size. T...
详细信息
ISBN:
(纸本)0780391543
Audio coding schemes such as MP3 or MPEG AAC can use a variable number of bits to encode each signal frame. However they can still be used oil constant bit rate channels thanks to a variation buffer limited in size. This buffer allows the encoder to use more bits for high complexity frames in order to maintain quality, the buffer being decreased e.g. for silent frames, leading to a constant rate coding on average. The use of such a buffer aims at reducing quality fluctuations over time, thanks to reasonable instantaneous bit rate variations. In this paper, a new design for a buffer controller is presented. This controller aims at finding the appropriate number of bits to encode each audio frame, according to chosen perceptual criteria and reduces tile perceived distortion over time. This process can take place in a two pass encoding process, the First pass aiming at measuring frames complexity and bit demand, the second consisting in allocating the appropriate number of bits for each. In order to cope with online coding application a single pass encoding with a reasonable additional delay is also investigated.
To overcome planning phases in spontaneous speech production, learners and native speakers use strategies such as (un)filled pauses, smallwords or discourse markers. Small scale studies in this vein have demonstrated ...
详细信息
To overcome planning phases in spontaneous speech production, learners and native speakers use strategies such as (un)filled pauses, smallwords or discourse markers. Small scale studies in this vein have demonstrated that learners differ from native speakers in that they underuse smallwords and discourse markers, and rely on other fluency-enhancing strategies instead. In the present paper, we present a corpus-based study, which investigates fluency-enhancing strategies in four components of the Louvain International Database of Spoken English Interlanguage (LINDSEI;Gilquin et al. 2010), covering four learner English varieties, namely Spanish, German, Bulgarian and Japanese. We investigate 216 different fluencemes (i.e. fluency-enhancing features;Gotz in Fluency in native and nonnative English speech, John Benjamins, Amsterdam, 2013) in 200 transcribed interviews with advanced learners of English. An online coding application, which was specially designed and programmed for this project, enables us to cover such a large amount of data. We report on the design, functionality and (dis-)advantages of the onlineapplication, the multilevel-coding system we implemented, and the methodological challenges we face in detail. We will also present the findings of one first pilot study where we exhibit considerable variation between and within learners of particular native languages concerning fluenceme frequencies, while distributional patterns of fluencemes are rather similar across varieties.
暂无评论