Contrast Enhancement at Borders Spatial Contrast Enhancement at Borders

meanings, we use context to select the meaning intended by the speaker. But context is involved at an even earlier level in speech perception because it allows us to recognize sounds as words. We perceive spoken words as a sequence of vowel and consonant sounds, that is, as a sequence of phonemes. Perceptually, each phoneme seems fixed, but acoustically, a phoneme can vary considerably depending on context. We hear the same /d/ sound in die and due, but the initial auditory signal of these words is different. This difference arises because of the flowing nature of speech production. As we articulate one phoneme, we are preparing to articulate the next, and the resulting co-articulation is reflected in the acoustical signal. The processes that underlie speech perception use contextual information to automatically and effortlessly compensate for the effect of co-articulation, resulting in a phonemic constancy similar to the lightness constancy described earlier. In some situations, such as a noisy party or a crowded restaurant, the acoustical signal may be insufficient for reliable speech recognition. In these situations, we may rely heavily on lipreading to understand what is being said. But even when the speech signal is clear and we are not consciously paying attention to the speaker’s mouth, visual context can influence phoneme perception. This automatic processing of facial movements explains why a dubbed movie can be particularly difficult to understand: We cannot ignore the visual context even when it is misleading. Like the other aspects of perception described here, phoneme recognition is a topic of intensive research. Although this process is not completely understood, a full description will involve the integration of contextual information across time and across senses. Although it may seem paradoxical, we perceive phonemes as having stable properties that are independent of context precisely because we are so adept at interpreting them in context.