The channel (or voice) vocoder operates as a bank of filters that breaks two incoming sound sources (the carrier and the modulator) into compatible frequency regions. The envelope within each subband of the modulator is imposed on the appropriate subband of the carrier, and the resulting sounds are added together. As shown below, the rectification nonlinearity followed by a lowpass filter approximates the envelope of the sound within the band. The channel vocoder can be used to generate a classic robotic-voice when modulated with speech, and it has found extensive use as a special effect in Hollywood. The channel vocoder is also often used to simulate the experience of hearing speech transduced by a cochlear implant. Please check for details: http://120years.net/the-voder-vocoderhomer-dudleyusa1940/ The channel vocoder can be interpreted as a filter-bank that imposes the envelope of one sound (the modulator) onto the waveform of another (the carrier). The process is also similar to amplitude modulation. The envelope operation (represented here by the application of a rectification nonlinearity g(x) followed by a lowpass filter) is applied separately within each frequency band.

Matlab implementation of the channel vocoder Tone-vocoded speech Tone-vocoded speech is created by dividing the speech signal into logarithmically-spaced frequency bands. In each frequency band, the amplitude envelope (half-wave or full-wave rectifier and a low-pass filter) is extracted and then used to modulate a sine wave carrier at the center of each band. Finally, each of the modulated bands is band-passed again (same band-pass filter), and then they are recombined (summation) to create the tone-vocoded sentence. (the band-pass filters for band-limiting are not required for tone vocoders; they are only necessary for noise vocoders) The example of a 6-band vocoder is as follows. A vocoded sound signal can be created by dividing the speech signal into six logarithmically-spaced frequency bands. The boundary and center frequencies of each band are shown in the table below, Frequency (/Hz) Band Min Center Max 1 50 140 229 2 229 394 558 3 558 860 1161 4 1161 1713 2265 5 2265 3278 4290 6 4290 6145 8000 For different settings, the following Matlab function can be called to generate the boundary frequencies of an N-band vocoder. The output vector contains N+1 frequencies with each adjacent pair for one band-pass filter. For example, Fco=[50 229 558 1161 2265 4290 8000] for N=6 as shown in the table above. Please set fmin to 300 Hz and fmax to 6000 Hz. You can vary the number of bands N. %----------------design of the logarithmic filter bank------------------------- % fmin: the lowest frequency to cover % fmax: the highest frequency to cover % N: number of bands or number of channels % The output fco contains N+1 frequency values for N bandpass filters. For example, [fco(i) fco(i+1)] is for the ith bandpass filter. function fco=bands_cutoff(fmin, fmax, N); xmin = log10(fmin/165.4+1)/2.1; xmax = log10(fmax/165.4+1)/2.1; %relative value dx = (xmax-xmin)/N; x = xmin:dx:xmax; fco=zeros(1,N+1); for i=1:N+1 fco(i)=165.4*(10^(x(i)*2.1)-1); end; The band-pass filter bank in Fig. 2 can be implemented with the Matlab command butter. Try to type in help butter to see how to use the command. Please set the order of each band-pass filter to 3 and the order of the low-pass filter to 2. You can use the command freqz to view the frequency response of a filter. The low-pass filters in Fig. 2 can also be implanted with the same butter command. Please set the low-pass cutoff frequency to 400 Hz for all the bands. You can also use the command freqz to view its frequency response. For example, the following codes demonstrate how you can design a band-pass filter from 1,000 Hz to 3,000 Hz and then plot its frequency response: Fs=16000; % sampling frequency [BB,AA]=butter(3,[1000 3000]/(Fs/2)); % The filter order is set to 3 and [1000 3000]/(Fs/2) are normalized frequencies between 0 and 1. [H,F]=freqz(BB,AA,256,Fs); % Fs is the sampling frequency plot(F,abs(H)); Then you can use the filter command in Matlab to filter the input signal. For example Y = filter(BB,AA,X); % X is the input and Y is the filtered signal. Type help butter in Matlab to learn how to design a band-pass or low-pass filter. The tone or sine carrier in Fig. 2 for each band can be generated with the following commands: t=(0:length(orig)-1)/Fs; % creating a time vector tone_carrier=sin(2*pi*Fc*t); The sampling frequency of the input .wav file is Fs and the center frequency of the corresponding band is at Fc. Here, orig means a vector containing the sound signal loaded by the audioread command in Matlab. Requirements: 1) Write your script file (e.g. using a for loop) in Matlab to implement the N-band tone vocoder as shown in Fig 2. The Matlab program should be able to process any sound file (.wav) you choose. The following code structure is for your reference only: % load a.wav file % call the bands_cutoff function to generate N+1 cutoff frequencies % initialize a vector with all zeros to store the sum of all bands. Its length is % the same as the sound signals length % for loop (i=1:N) % Design the ith band-pass filter (BPF) % apply the BPF % rectification % design a low-pass filter (LPF) % apply the LPF to the rectified signal to extract its envelope % calculate the center frequency of the ith band % create a time vector (same length as the sound signals) % generate a sine wave at the center frequency of the ith band % multiplication (you may have to transpose one of them if you % receive an error Dimensions do not match ) % update the sum at each iteration to get the sum of all bands % end of the loop % Normalize the signal % Play the sound signal %

2) The script file should be able to load a .wav sound file and process it to generate a vocoded sound. 3) Try to process different types of .wav sound files (such as a music signal) that you can download from the internet. Make sure all the filters you designed to use the actual sampling frequency of your sound signal. 4) Be able to show some of the intermediate waveforms in the vocoder processing, such as band-passed signals, envelopes extracted (after rectification and low-pass filtering), and waveforms of the original sound and vocoded sound. 5) Show demos of your vocoded sounds to demonstrate that you have correctly implemented the vocoder. Make sure your sound signals are normalized to [-1, +1] when you use the Matlab command sound to play them.