tomwalters@32: #!/usr/bin/env python tomwalters@32: # encoding: utf-8 tomwalters@32: # tomwalters@32: # AIM-C: A C++ implementation of the Auditory Image Model tomwalters@32: # http://www.acousticscale.org/AIMC tomwalters@32: # tomwalters@45: # Licensed under the Apache License, Version 2.0 (the "License"); tomwalters@45: # you may not use this file except in compliance with the License. tomwalters@45: # You may obtain a copy of the License at tomwalters@32: # tomwalters@45: # http://www.apache.org/licenses/LICENSE-2.0 tomwalters@32: # tomwalters@45: # Unless required by applicable law or agreed to in writing, software tomwalters@45: # distributed under the License is distributed on an "AS IS" BASIS, tomwalters@45: # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. tomwalters@45: # See the License for the specific language governing permissions and tomwalters@45: # limitations under the License. tomwalters@32: """ tomwalters@32: Profiles_test.py tomwalters@32: tomwalters@32: Created by Thomas Walters on 2010-02-22. tomwalters@32: Copyright 2010 Thomas Walters tomwalters@32: Test the AIM-C model from filterbank to SSI profiles tomwalters@32: """ tomwalters@32: tomwalters@32: import aimc tomwalters@32: from scipy.io import wavfile tomwalters@32: from scipy import io tomwalters@32: import scipy tomwalters@32: import pylab tomwalters@32: from itertools import izip, chain, repeat tomwalters@32: tomwalters@32: def grouper(n, iterable, padvalue=None): tomwalters@32: "grouper(3, 'abcdefg', 'x') --> ('a','b','c'), ('d','e','f'), ('g','x','x')" tomwalters@32: return izip(*[chain(iterable, repeat(padvalue, n-1))]*n) tomwalters@32: tomwalters@32: def BankToArray(out_bank): tomwalters@32: channel_count = out_bank.channel_count() tomwalters@32: out_buffer_length = out_bank.buffer_length() tomwalters@32: out = scipy.zeros((channel_count, out_buffer_length)) tomwalters@32: for ch in range(0, channel_count): tomwalters@32: for i in range(0, out_buffer_length): tomwalters@32: out[ch, i] = out_bank.sample(ch, i) tomwalters@32: return out tomwalters@32: tomwalters@32: def StrobesToList(bank): tomwalters@32: channel_count = bank.channel_count() tomwalters@32: strobes = [] tomwalters@32: for ch in range(0, channel_count): tomwalters@32: s = [] tomwalters@32: for i in range(0, bank.strobe_count(ch)): tomwalters@32: s.append(bank.strobe(ch, i)) tomwalters@32: strobes.append(s) tomwalters@32: tomwalters@32: def main(): tomwalters@32: wave_path = "/Users/Tom/Documents/Work/PhD/HTK-AIM/Sounds/" tomwalters@32: #features_path = "/Users/Tom/Documents/Work/PhD/HTK-AIM/work08-jess-original-rec_rubber/features/" tomwalters@32: tomwalters@32: file_name = "ii/ii172.5p112.5s100.0t+000itd" tomwalters@32: tomwalters@32: wave_suffix = ".wav" tomwalters@32: features_suffix = ".mat" tomwalters@32: tomwalters@32: frame_period_ms = 10; tomwalters@32: tomwalters@32: wave_filename = wave_path + file_name + wave_suffix tomwalters@32: #features_filename = features_path + file_name + features_suffix tomwalters@32: tomwalters@32: (sample_rate, input_wave) = wavfile.read(wave_filename) tomwalters@32: wave_length = input_wave.size tomwalters@32: buffer_length = int(frame_period_ms * sample_rate / 1000) tomwalters@32: tomwalters@32: #pylab.plot(input_wave) tomwalters@32: #pylab.show() tomwalters@32: tomwalters@32: input_sig = aimc.SignalBank() tomwalters@32: input_sig.Initialize(1, buffer_length, sample_rate) tomwalters@32: parameters = aimc.Parameters() tomwalters@32: parameters.SetFloat("sai.frame_period_ms", 10.0) tomwalters@32: parameters.SetInt("input.buffersize", 480) tomwalters@32: tomwalters@32: mod_gt = aimc.ModuleGammatone(parameters) tomwalters@32: mod_hl = aimc.ModuleHCL(parameters) tomwalters@32: mod_strobes = aimc.ModuleLocalMax(parameters) tomwalters@32: mod_sai = aimc.ModuleSAI(parameters) tomwalters@32: parameters.SetBool("ssi.pitch_cutoff", True) tomwalters@32: parameters.SetBool("ssi.weight_by_cutoff", False) tomwalters@32: parameters.SetBool("ssi.weight_by_scaling", True) tomwalters@32: parameters.SetBool("ssi.log_cycles_axis", True) tomwalters@32: mod_ssi = aimc.ModuleSSI(parameters) tomwalters@32: tomwalters@32: parameters.SetFloat("nap.lowpass_cutoff", 100.0) tomwalters@32: mod_nap_smooth = aimc.ModuleHCL(parameters) tomwalters@32: mod_scaler = aimc.ModuleScaler(parameters) tomwalters@32: tomwalters@32: parameters.SetBool("slice.all", False) tomwalters@32: parameters.SetInt("slice.lower_index", 77) tomwalters@32: parameters.SetInt("slice.upper_index", 150) tomwalters@32: slice_1 = aimc.ModuleSlice(parameters) tomwalters@32: tomwalters@32: parameters.SetInt("slice.lower_index", 210) tomwalters@32: parameters.SetInt("slice.upper_index", 240) tomwalters@32: slice_2 = aimc.ModuleSlice(parameters) tomwalters@32: tomwalters@32: parameters.SetInt("slice.lower_index", 280) tomwalters@32: parameters.SetInt("slice.upper_index", 304) tomwalters@32: slice_3 = aimc.ModuleSlice(parameters) tomwalters@32: tomwalters@32: parameters.SetInt("slice.lower_index", 328) tomwalters@32: parameters.SetInt("slice.upper_index", 352) tomwalters@32: slice_4 = aimc.ModuleSlice(parameters) tomwalters@32: tomwalters@32: parameters.SetBool("slice.all", True) tomwalters@32: slice_5 = aimc.ModuleSlice(parameters) tomwalters@32: tomwalters@32: nap_profile = aimc.ModuleSlice(parameters) tomwalters@32: tomwalters@32: features_1 = aimc.ModuleGaussians(parameters) tomwalters@32: features_2 = aimc.ModuleGaussians(parameters) tomwalters@32: features_3 = aimc.ModuleGaussians(parameters) tomwalters@32: features_4 = aimc.ModuleGaussians(parameters) tomwalters@32: features_5 = aimc.ModuleGaussians(parameters) tomwalters@32: tomwalters@32: mod_gt.AddTarget(mod_hl) tomwalters@32: mod_gt.AddTarget(mod_nap_smooth) tomwalters@32: mod_nap_smooth.AddTarget(nap_profile) tomwalters@32: nap_profile.AddTarget(mod_scaler) tomwalters@32: mod_hl.AddTarget(mod_strobes) tomwalters@32: mod_strobes.AddTarget(mod_sai) tomwalters@32: mod_sai.AddTarget(mod_ssi) tomwalters@32: mod_ssi.AddTarget(slice_1) tomwalters@32: mod_ssi.AddTarget(slice_2) tomwalters@32: mod_ssi.AddTarget(slice_3) tomwalters@32: mod_ssi.AddTarget(slice_4) tomwalters@32: mod_ssi.AddTarget(slice_5) tomwalters@32: tomwalters@32: slice_1.AddTarget(features_1) tomwalters@32: slice_2.AddTarget(features_2) tomwalters@32: slice_3.AddTarget(features_3) tomwalters@32: slice_4.AddTarget(features_4) tomwalters@32: slice_5.AddTarget(features_5) tomwalters@32: tomwalters@32: mod_gt.Initialize(input_sig) tomwalters@32: tomwalters@32: correct_count = 0; tomwalters@32: incorrect_count = 0; tomwalters@32: tomwalters@32: scaled_wave = [] tomwalters@32: for sample in input_wave: tomwalters@32: scaled_wave.append(float(sample / float(pow(2,15) - 1))) tomwalters@32: i = 0 tomwalters@32: tomwalters@32: wave_chunks = grouper(buffer_length, scaled_wave, 0) tomwalters@32: tomwalters@32: out_bmm = [] tomwalters@32: out_nap = [] tomwalters@32: out_smooth_nap_profile = [] tomwalters@32: out_strobes = [] tomwalters@32: out_sais = [] tomwalters@32: out_ssis = [] tomwalters@32: out_slice_1 = [] tomwalters@32: out_slice_2 = [] tomwalters@32: out_slice_3 = [] tomwalters@32: out_slice_4 = [] tomwalters@32: out_slice_5 = [] tomwalters@32: out_feat_1 = [] tomwalters@32: out_feat_2 = [] tomwalters@32: out_feat_3 = [] tomwalters@32: out_feat_4 = [] tomwalters@32: out_feat_5 = [] tomwalters@32: for chunk in wave_chunks: tomwalters@32: i = 0 tomwalters@32: for sample in chunk: tomwalters@32: input_sig.set_sample(0, i, float(sample)) tomwalters@32: i += 1 tomwalters@32: mod_gt.Process(input_sig) tomwalters@32: tomwalters@32: #out_bmm.append(BankToArray(mod_gt.GetOutputBank())) tomwalters@32: #out_nap.append(BankToArray(mod_hl.GetOutputBank())) tomwalters@32: out_smooth_nap_profile.append(BankToArray(mod_scaler.GetOutputBank())) tomwalters@32: #out_strobes.append(BankToArray(mod_strobes.GetOutputBank())) tomwalters@32: #out_sais.append(BankToArray(mod_sai.GetOutputBank())) tomwalters@32: out_ssis.append(BankToArray(mod_ssi.GetOutputBank())) tomwalters@32: out_slice_1.append(BankToArray(slice_1.GetOutputBank())) tomwalters@32: out_slice_2.append(BankToArray(slice_2.GetOutputBank())) tomwalters@32: out_slice_3.append(BankToArray(slice_3.GetOutputBank())) tomwalters@32: out_slice_4.append(BankToArray(slice_4.GetOutputBank())) tomwalters@32: out_slice_5.append(BankToArray(slice_5.GetOutputBank())) tomwalters@32: out_feat_1.append(BankToArray(features_1.GetOutputBank())) tomwalters@32: out_feat_2.append(BankToArray(features_2.GetOutputBank())) tomwalters@32: out_feat_3.append(BankToArray(features_3.GetOutputBank())) tomwalters@32: out_feat_4.append(BankToArray(features_4.GetOutputBank())) tomwalters@32: out_feat_5.append(BankToArray(features_5.GetOutputBank())) tomwalters@32: tomwalters@32: out_bank = mod_gt.GetOutputBank() tomwalters@32: channel_count = out_bank.channel_count() tomwalters@32: cfs = scipy.zeros((channel_count)) tomwalters@32: for ch in range(0, channel_count): tomwalters@32: cfs[ch] = out_bank.centre_frequency(ch) tomwalters@32: outmat = dict(bmm=out_bmm, nap=out_nap, sais=out_sais, tomwalters@32: ssis=out_ssis, slice1=out_slice_1, slice2=out_slice_2, tomwalters@32: slice3=out_slice_3, slice4=out_slice_4, slice5=out_slice_5, tomwalters@32: feat1=out_feat_1, feat2=out_feat_2, feat3=out_feat_3, tomwalters@32: feat4=out_feat_4, feat5=out_feat_5, tomwalters@32: nap_smooth=out_smooth_nap_profile, centre_freqs=cfs) tomwalters@32: io.savemat("src/Scripts/profile_out.mat", outmat, oned_as='column') tomwalters@32: tomwalters@32: pass tomwalters@32: tomwalters@32: tomwalters@32: if __name__ == '__main__': tomwalters@32: main()