tom@516: % Copyright 2012, Google, Inc.
tom@516: % Author: Richard F. Lyon
tom@516: %
tom@516: % This Matlab file is part of an implementation of Lyon's cochlear model:
tom@516: % "Cascade of Asymmetric Resonators with Fast-Acting Compression"
tom@516: % to supplement Lyon's upcoming book "Human and Machine Hearing"
tom@516: %
tom@516: % Licensed under the Apache License, Version 2.0 (the "License");
tom@516: % you may not use this file except in compliance with the License.
tom@516: % You may obtain a copy of the License at
tom@516: %
tom@516: %     http://www.apache.org/licenses/LICENSE-2.0
tom@516: %
tom@516: % Unless required by applicable law or agreed to in writing, software
tom@516: % distributed under the License is distributed on an "AS IS" BASIS,
tom@516: % WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
tom@516: % See the License for the specific language governing permissions and
tom@516: % limitations under the License.
tom@516: 
tom@516: function state = CARFAC_AGCStep(AGC_coeffs, avg_detects, state)
tom@516: % function state = CARFAC_AGCStep(AGC_coeffs, avg_detects, state)
tom@516: %
tom@516: % one time step (at decimated low AGC rate) of the AGC state update
tom@516: 
tom@516: n_AGC_stages = length(AGC_coeffs.AGC_epsilon);
tom@516: n_mics = length(state);
tom@516: n_ch = size(state(1).AGC_sum, 1);  % number of channels
tom@516: 
tom@516: optimize_for_mono = n_mics == 1;  % mono optimization
tom@516: if ~optimize_for_mono
tom@516:   stage_sum = zeros(n_ch, 1);
tom@516: end
tom@516: 
tom@516: for stage = 1:n_AGC_stages
tom@516:   if ~optimize_for_mono  % skip if mono
tom@516:     if stage > 1
tom@516:       prev_stage_mean = stage_sum / n_mics;
tom@516:     end
tom@516:     stage_sum(:) = 0;  % sum accumulating over mics at this stage
tom@516:   end
tom@516:   epsilon = AGC_coeffs.AGC_epsilon(stage);  % for this stage's LPF pole
tom@516:   polez1 = AGC_coeffs.AGC1_polez(stage);
tom@516:   polez2 = AGC_coeffs.AGC2_polez(stage);
tom@516:   for mic = 1:n_mics
tom@516:     if stage == 1
tom@516:       AGC_in = AGC_coeffs.detect_scale * avg_detects(:,mic);
tom@516:       AGC_in = max(0, AGC_in);  % don't let neg inputs in
tom@516:     else
tom@516:       % prev. stage mixed with prev_stage_sum
tom@516:       if optimize_for_mono
tom@516:         % Mono optimization ignores AGC_mix_coeff,
tom@516:         % assuming all(prev_stage_mean == AGC_memory(:, stage - 1));
tom@516:         % but we also don't even allocate or compute the sum or mean.
tom@516:         AGC_in = AGC_coeffs.AGC_stage_gain * ...
tom@516:           state(mic).AGC_memory(:, stage - 1);
tom@516:       else
tom@516:         AGC_in = AGC_coeffs.AGC_stage_gain * ...
tom@516:           (AGC_coeffs.AGC_mix_coeff * prev_stage_mean + ...
tom@516:             (1 - AGC_coeffs.AGC_mix_coeff) * ...
tom@516:               state(mic).AGC_memory(:, stage - 1));
tom@516:       end
tom@516:     end
tom@516:     AGC_stage = state(mic).AGC_memory(:, stage);
tom@516:     % first-order recursive smooting filter update:
tom@516:     AGC_stage = AGC_stage + epsilon * (AGC_in - AGC_stage);
tom@516: 
tom@516:     % spatially spread it; using diffusion coeffs like in smooth1d
tom@516:     AGC_stage = SmoothDoubleExponential(AGC_stage, polez1, polez2);
tom@516: 
tom@516:     state(mic).AGC_memory(:, stage) = AGC_stage;
tom@516:     if stage == 1
tom@516:       state(mic).sum_AGC = AGC_stage;
tom@516:     else
tom@516:       state(mic).sum_AGC = state(mic).sum_AGC + AGC_stage;
tom@516:     end
tom@516:     if ~optimize_for_mono
tom@516:       stage_sum = stage_sum + AGC_stage;
tom@516:     end
tom@516:   end
tom@516: end
tom@516: 
tom@516: