Applied Speech and Audio Processing: With matlab examples
Download 2.66 Mb. Pdf ko'rish
|
Applied Speech and Audio Processing With MATLAB Examples ( PDFDrive )
Advanced topics
further improvement, Festival attempts to speak syllables rhythmically, adjust speaking duration, modify pitch to fit accents and any tonality which may be required. Users may also inform the synthesiser concerning the class of what is being spoken – such as words, phrases, raw text, and so on, and provide relevant information to specify phrasing, timings and even tone changes [22]. 7.7 Stereo encoding Stereo means something that is stored or reproduced in more than one channel. It can also be referred to as stereophonic. This was contrasted to the technical term ‘binaural’ in Section 4.2.13 which means something pertaining to both ears. In audio terms, stereo normally refers to two channels of audio, one for a right loudspeaker and one for a left loudspeaker, but many other multichannel arrangements exist. The basic idea of stereo audio relates to the fact that most of us have two ears which we use to differentiate between sounds, particularly regarding source placement. In nature, a sound to the left of our heads will be heard by the left ear slightly before it is heard by the right ear, and slightly louder. This tiny delay, or phase difference, between the two versions of sound heard by the ears is sufficient to allow the brain to calculate where the sound originated from. We are not conscious of this calculation since it is performed automatically, but we can demonstrate the effect very easily using Matlab. First we will create a sound made from two related sinewaves. These need to be sampled at a higher frequency than some of our demonstrations because of the more fine control over timing we can achieve. In this case we will use a 44.1 kHz sample rate to create a 440 Hz note in the usual way: Fs=44100; Ft=440; note=tonegen(Ft, Fs, 0.5) Next we will replay two versions of this using soundsc. In Matlab, stereo sound is represented by a two-column matrix of samples. The first column holds samples for the left channel, and the second column holds samples for the right channel. So we will create two two-column matrices of audio, one in which the left channel is delayed slightly, and one in which the right channel is delayed slightly: s1=[[zeros(1,20), note];[note, zeros(1,20)]]; s2=fliplr(s1); where fliplr() is used to switch the left and right channels around. Next we will listen to those two sounds. You will definitely need a computer capable of replaying in stereo (and most are these days), and quite likely will require a pair of stereo headphones to really appreciate the stereo placement: |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling