CWT design in simulink - (mathworks.com) thread is the answer. .I use the Discrete FIR Filter block on interval N where instead of coefficients you put the wavelet represented by N samples. The block will perform cross-correlation of you signal N samples on one input, to the wavelet on another input at the interval N. By using different wavelets you can determine then, which wavelet represent signal better in form of immediate magnitude. By performing this continuosly as the signal evolves (shifts) sample by sample you will get cwt() on that evolvement interval. To facilitate the shift register of length N I use the DSP toolbox Buffer block, where I set the buffer length to N, and buffer overlap to N-1, so size of the Buffer output will be the same as your wavelet length.
So each sample time at the output of the Discrete FIR Filter you will be getting the output in shape of the vector , the size of which is equal to the amount of your wavelets, or frequencies, since each wavelet is tuned to particular frequency. Each member of the vector is the Sum of N products between samples of the signal and the wavelet at the input of Discrete FIR Filter. The output vector will be representing a single sample of Continuous Wavelet Transform ,which over time will draw a picture like below, which I borrowed from Mathworks cwt() help examples.
