Applied Speech and Audio Processing: With matlab examples
Download 2.66 Mb. Pdf ko'rish
|
Applied Speech and Audio Processing With MATLAB Examples ( PDFDrive )
Hearing
4.3 Amplitude and frequency models Attempts to classify and regularise the human hearing system have been made for many years. Most of these have revolved around the analysis of frequency and amplitude effects, and in particular the measurements of loudness, and frequency selectivity dis- cussed below. 4.3.1 Loudness Backtracking from the equal loudness contours of Section 4.2.1, it is fairly evident that humans do not measure the concept of loudness in the same way as a physical mea- surement. From our discussions previously we have seen that people interpret amplitude approximately logarithmically. In fact, several researchers indicate that perceived loudness is proportional to physical amplitude raised to the power of 0.3 [4]. It is likely that there are upper and lower bounds where the relationship no longer holds true, nevertheless across the main portion of the hearing range the relationship is defined by a constant multiplier that varies from person to person. The measure of loudness is the sone although this is rarely used in current literature. One sone is the loudness of a 1 kHz tone at 40 dB SPL , with every 10 dB increase in amplitude increasing the loudness by one sone (i.e. a 50 dB SPL 1 kHz tone has a loudness of two sones). 4.3.2 The Bark scale Since the hearing process is often considered to derive from some set of bandpass fil- ters, and the resultant processing within these, researchers attempted to identify and characterise these filters [4,10,11]. Known as critical band filters, each has similar shape but different bandwidth, centre frequency, and amplitude weighting. The different ampli- tude weightings contribute to the sone loudness scale (Section 4.3.1), as well as the equal loudness contours of Section 4.2.1 – both of which originally described the loudness of single tones only. More generally, the loudness and frequency selectivity of sounds depend on the bands within which they fall. In particular this mechanism can be used to explain the masking effects of Section 4.2.8: a loud tone in one critical band will occupy that band and prevent a quieter tone in the same band from being heard. Moving the quieter tone gradually away from the louder tone until it enters a neighbouring critical band will result in it becoming audible once more. The Bark 1 scale is one way to express this relationship [20]. The first few lower critical bands are shown in Table 4.3 where they are num- bered according to the Bark scale. The scale is arranged so that a unit change in Bark 1 Since the scale is named after a person, Heinrich Barkhausen, for his early work on loudness perception, the name Bark should be capitalised to distinguish it from the sound a dog makes. |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling