OSCR

Next-Generation Neural Mass Models Reproduce Features of Speech Processing.

Code ↔ Paper

21 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 21 matches · 1 of them tie a paragraph to a whole file, not to given lines: a weak match, whose lines are not tinted
  1. [1] § Materials and Methods › Model dynamics investigation › IEPC about phoneme onsets ↔ Plot_IEPC_ringing_after_end_of_stimulus.jl, lines 98–180 · score 0.78 · band pass filtered, Hz band, Hilbert transform, frequency band, argument, Butterworth
  2. [2] § Materials and Methods › Sharpness-specific tuning of tracking › Stimulus construction ↔ Generate_Noise.jl, the whole file · a weak match · score 0.77 · 80–8000 Hz, band pass filter, Hilbert transform, cochlear, summed, 80 Hz
  3. [3] § Materials and Methods › Evaluation metrics and statistical analyses › Inter event phase coherence ↔ src/Phase_Concentration_Metric.jl, lines 231–295 · score 0.74 · analytic signal, filtered signals, complex unit vectors, instantaneous phase, Hilbert, phoneme
  4. [4] § Materials and Methods › Evaluation metrics and statistical analyses › Correlation and multiple comparison corrections ↔ Plot_paper_results_figures_and_stats.jl, lines 456–498 · score 0.73 · Confidence intervals, Benjamini Hochberg, Bonferroni correction, tailed, FDR, correlation
  5. [5] § Materials and Methods › Evaluation metrics and statistical analyses › Phase concentration metric ↔ src/Phase_Concentration_Metric.jl, lines 231–295 · score 0.71 · filtered signal, complex unit vectors, instantaneous phases, kernel, Hilbert, metric
  6. [6] § Materials and Methods › Evaluation metrics and statistical analyses › Inter event phase coherence ↔ Plot_IEPC_ringing_after_end_of_stimulus.jl, lines 98–180 · score 0.70 · band pass filtered, Hilbert transformed, frequency band, circular, IEPC, windows
  7. [7] § Materials and Methods › Model dynamics investigation › IEPC about phoneme onsets ↔ Plot_IEPC.jl, lines 118–163 · score 0.70 · Hilbert transform, frequency band, 0.67 Hz, Butterworth, 35 Hz, filtered
  8. [8] § Materials and Methods › Sharpness-specific tuning of tracking › Stimulus construction ↔ src/NGNMM_NSP_paper_code.jl, lines 1389–1445 · score 0.67 · peak rate impulse, peak magnitude, low pass filter, threshold, events, smoothed
  9. [9] § Materials and Methods › Model dynamics investigation › PCM experiment ↔ src/Phase_Concentration_Metric.jl, lines 1–70 · score 0.64 · generated sine wave, stimulus frequency, metrics, PCM, impulses, peaks
  10. [10] § Materials and Methods › Model dynamics investigation › PCM experiment ↔ src/Phase_Concentration_Metric.jl, lines 297–350 · score 0.62 · Gaussian filtered, unit vectors, model response, metrics, PCM, evoked
  11. [11] § Materials and Methods › Sharpness-specific tuning of tracking › Stimulus construction ↔ src/NGNMM_NSP_paper_code.jl, lines 262–347 · score 0.57 · noise scaled, concatenated, squeezing, stretching, stimulus envelopes, ratio
  12. [12] § Materials and Methods › Sharpness-specific tuning of tracking › Stimulus construction ↔ Plot_IEPC.jl, lines 1–65 · score 0.56 · peak rate impulse, peak magnitude, threshold, events, smoothed, derivative
  13. [13] § Results ↔ Compute_and_plot_PCM.jl, lines 97–185 · score 0.56 · tuned phase resetting, sine waves, fast NMM, slow NMM, phase resetting oscillator, arrows
  14. [14] § Materials and Methods › Evaluation metrics and statistical analyses › Concordance correlation coefficient ↔ src/NGNMM_NSP_paper_code.jl, lines 1490–1532 · score 0.56 · concordance correlation coefficient, model ITPCs, variances, noise
  15. [15] § Materials and Methods › Model dynamics investigation › PCM experiment ↔ src/NGNMM_NSP_paper_code.jl, lines 93–146 · score 0.55 · strictly positive, sine wave, drive amplitude, PCM, phase resetting, oscillator
  16. [16] § Results ↔ src/Phase_Concentration_Metric.jl, lines 1–70 · score 0.55 · Phase concentration, sine waves, stimulus frequency, impulses, vectors, PCM
  17. [17] § Results ↔ src/Phase_Concentration_Metric.jl, lines 72–132 · score 0.55 · impulse train, oscillator response, peak rate, Error, event, derivative
  18. [18] § Results ↔ Plot_IEPC.jl, lines 165–249 · score 0.53 · transient response, phoneme onset, subtracted, IEPC, baseline, window
  19. [19] § Materials and Methods › Evaluation metrics and statistical analyses › Phase concentration metric ↔ src/Phase_Concentration_Metric.jl, lines 205–229 · score 0.53 · frequency domain, Gaussian, width, metric, concentration, filtered
  20. [20] § Results ↔ Plot_paper_results_figures_and_stats.jl, lines 692–728 · score 0.52 · unvoiced stops, nasals, sibilants, EEG, vowel, ITPC
  21. [21] § Materials and Methods › Model dynamics investigation › IEPC about phoneme onsets ↔ Plot_IEPC.jl, lines 165–249 · score 0.51 · phoneme onset, subtracted, segmented, IEPC, baseline, transient

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

Julia · 403 lines · 22 KB · MIT · 7 matches

  1. ### code to run Phase concentration metric tests on each model.
  2. ### needs sine wave stimuli, from which impulse trains of a range of frequencies can be extracted from the peaks.
  4. using FFTW, DSP, Statistics
  7. ### reformed drive interpolators generator to create sine waves of varying frequency for PCM tests.
  8. """
  9. Will make an "interpolators" array that contains interpolations of 20 sine waves of different frequency and put it in global variable Interpolators.
  10. This can then be used in the ensemble simulation via the interpolation selector variable in a prob_func
  11. """
  12. function generate_sine_drive_interpolators_for_saving(sampling_rate::Float64,frequency_range::Tuple{Float64,Float64},stimulus_set_number::Int)
  13. sr=sampling_rate
  14. stimlength=sr*10 #code expects 10 seconds, 5 seconds of noise, 5 seconds of stimulus.
  15. times=range(0.0,10.0,length=Int(stimlength))
  16. frequencies=range(frequency_range[1],frequency_range[2],length=20*3) #20 frequencies per stimulus set, 3 stimulus sets. 60 freqs.
  17. sine_drives=Vector{Any}(undef,20)
  19. for trial in 1:20
  20. #will go through the noise samples in triplets: 1 is the before stimulus noise, 1 is during stimulus and 1 is after stimulus
  21. #will loop over this, and create a single interpolator for each set of noise samples, with the stimulus added to the middle and appropriate scale applied
  22. #drive is [noise,stimulus,noise,zeroes] (in case it runs over, making extrapolation to be 0 only)
  23. trial_drive_input=sin.(2.0*pi.*frequencies[trial+20*(stimulus_set_number-1)].*times) #only for the appropriate 20 frequencies of this stimulus set
  25. xs=1:1:length(trial_drive_input)
  27. sine_drives[trial]=linear_interpolation(xs,trial_drive_input,extrapolation_bc=Line())
  28. #interpolators[trial]=interpolate(Vector(trial_drive_input),BSpline(Linear()))#,extrapolation_bc=Line())
  29. end
  30. return sine_drives #look into typed global
  31. end
  33. # new prob func for the oscillator. this will update its frequency to match the stimulus frequency.
  34. function vary_noise_and_initial_conditions_and_natfreq(prob,i,repeat)
  35. prob.p.noise_selector=i
  36. freqs=prob.p.frequencies
  37. prob.p.F=freqs[Int64(i+20*(prob.p.noise_case_reference-1))] #set natural frequency to match stimulus frequency.
  38. prob.u0[1]=rand(MersenneTwister(Int64(123+i+20*(prob.p.noise_case_reference-1))))*2*π-π #random phase, but keeping r=1.0. indexed by noise selector i = 1:1:60.
  39. prob
  40. end
  42. function get_coupled_oscillator_PCM_across_60freqs_randinitcond(p,u_init,time_range,PCMrange,freq_range,save_traj,savepath,stimulus_type)
  43. condition="PCM"
  44. println("testing: ",condition)
  45. flush(stdout)
  46. saveat=1/44100
  47. response_sr=Int64(1/saveat)
  48. frequencies=range(freq_range[1],freq_range[2],length=60)#[1:50] #cut to 50 to avoid failed simulations at high freq end.
  49. all_60_trials=Vector{Any}()
  50. all_stimuli=Vector{Any}()
  51. for stimulus_idx in 1:3 #3 sets of 20 stimuli to cover 60 frequencies, structured this way to match prior phoneme drive generation.
  52. println("stimulus set: ",stimulus_idx)
  53. p.noise_case_reference=stimulus_idx
  54. p.noise_selector=1
  55. flush(stdout)
  56. #set stimulus based on type: either normal envelopes, or derivative or event based impulse trains (peak rate or peak envelope)
  57. if stimulus_type=="envelope"
  58. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  59. elseif stimulus_type=="derivative"
  60. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  61. interpolators_global=get_smooth_derivative(interpolators_global,Float64(response_sr))
  62. elseif stimulus_type=="peakrate"
  63. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  64. interpolators_global=get_scaled_peakrate_impulses(interpolators_global,Float64(response_sr))
  65. elseif stimulus_type=="peakenvelope"
  66. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  67. interpolators_global=get_unitary_peakenv_impulses(interpolators_global, Float64(response_sr))
  68. else
  69. error("Unknown stimulus type: $(stimulus_type), must be one of 'envelope', 'derivative', 'peakrate', or 'peakenvelope'.")
  70. end
  72. #drive scaling as per phoneme case workflow.
  73. max_stimulus_amplitude=maximum(maximum.(interpolators_global))
  74. p.drive_amplitude=20000/mean(sum.([i[5.0*response_sr:10.0*response_sr] for i in interpolators_global]))
  75. p.c=1.0*pi*(1/(max_stimulus_amplitude*p.drive_amplitude*p.q)) #including q. this is the final maximumum stimulus amplitude after all normalisation.
  77. println("running sim")
  78. flush(stdout)
  79. if stimulus_idx==3 #to match the cut to 50 frequencies above. #no longer cut.
  80. trialdata=Ensemble_CoupledOscillators_modulated_PCM(vary_noise_and_initial_conditions_and_natfreq,time_range,p,u_init,20,saveat)
  81. else
  82. trialdata=Ensemble_CoupledOscillators_modulated_PCM(vary_noise_and_initial_conditions_and_natfreq,time_range,p,u_init,20,saveat)
  83. end
  84. push!(all_60_trials,trialdata...)
  85. push!(all_stimuli,interpolators_global...)
  86. end
  88. PCM_vectors,rates=calculate_PCM_oscillator_noisyrates(all_60_trials,all_stimuli,frequencies,response_sr,time_range,PCMrange) #ITPCrange can now exclude first 500ms of stimulus as per O.Cucu paper.
  89. #save the rates for each trial, for comparison across natural frequencies, and computation of conglomerate activity if desired.
  90. if save_traj
  91. #check dir exists:
  92. if !isdir(savepath*"Trajectories/")
  93. mkdir(savepath*"Trajectories/")
  94. end
  95. local name="Trajectories/$(response_sr)sr_oscillator_response_to_sine_stimulus_freq_range_$(freq_range[1])to$(freq_range[2])Hz_$(stimulus_type)_transformed"
  96. CSV.write(savepath*name*".csv",DataFrame(rates,:auto),writeheader=true)
  97. generate_arrow(name,savepath)
  98. rm(savepath*name*".csv")
  99. end
  100. return(PCM_vectors)
  101. end
  104. function get_evoked_model_PCM_across_60freqs_randinitcond(p,u_init,time_range,PCMrange,freq_range,save_traj,savepath,stimulus_type)
  105. condition="PCM"
  106. println("testing: ",condition)
  107. flush(stdout)
  108. saveat=1/44100
  109. response_sr=Int64(1/saveat)
  110. frequencies=range(freq_range[1],freq_range[2],length=60) #cut to 50 to avoid failed simulations at high freq end.
  111. all_60_trials=Vector{Any}()
  112. all_stimuli=Vector{Any}()
  113. for stimulus_idx in 1:3 #3 sets of 20 stimuli to cover 60 frequencies, structured this way to match prior phoneme drive generation.
  114. println("stimulus set: ",stimulus_idx)
  115. p.noise_case_reference=stimulus_idx
  116. p.noise_selector=1
  117. flush(stdout)
  118. #set stimulus based on type: either normal envelopes, or derivative or event based impulse trains (peak rate or peak envelope)
  119. if stimulus_type=="envelope"
  120. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  121. elseif stimulus_type=="derivative"
  122. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  123. interpolators_global=get_smooth_derivative(interpolators_global,Float64(response_sr))
  124. elseif stimulus_type=="peakrate"
  125. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  126. interpolators_global=get_scaled_peakrate_impulses(interpolators_global,Float64(response_sr))
  127. elseif stimulus_type=="peakenvelope"
  128. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  129. interpolators_global=get_unitary_peakenv_impulses(interpolators_global,Float64(response_sr))
  130. else
  131. error("Unknown stimulus type: $(stimulus_type), must be one of 'envelope', 'derivative', 'peakrate', or 'peakenvelope'.")
  132. end
  134. println("running sim")
  135. flush(stdout)
  136. if stimulus_idx==3 #to match the cut to 50 frequencies above. #no longer cut.
  137. trialdata=Ensemble_EvokedModel(vary_noise_and_initial_conditions_evokedmodel,time_range,p,u_init,20,saveat)
  138. else
  139. trialdata=Ensemble_EvokedModel(vary_noise_and_initial_conditions_evokedmodel,time_range,p,u_init,20,saveat)
  140. end
  141. push!(all_60_trials,trialdata...)
  142. push!(all_stimuli,interpolators_global...)
  143. end
  145. PCM_vectors,rates=calculate_PCM_EvokedModel_noisyrates(all_60_trials,all_stimuli,frequencies,response_sr,time_range,PCMrange) #ITPCrange can now exclude first 500ms of stimulus as per O.Cucu paper.
  146. #save the rates for each trial, for comparison across natural frequencies, and computation of congolmerate activity if desired.
  147. if save_traj
  148. #check dir exists:
  149. if !isdir(savepath*"Trajectories/")
  150. mkdir(savepath*"Trajectories/")
  151. end
  152. local name="Trajectories/evoked_model_response_to_sine_stimulus_freq_range_$(freq_range[1])to$(freq_range[2])Hz_$(stimulus_type)_transformed"
  153. CSV.write(savepath*name*".csv",DataFrame(rates,:auto),writeheader=true)
  154. generate_arrow(name,savepath)
  155. rm(savepath*name*".csv")
  156. end
  157. return(PCM_vectors)
  158. end
  160. function get_NGNMM_PCM_across_60freqs_randinitcond(p,u_init,time_range,PCMrange,freq_range,save_traj,savepath,stimulus_type)
  161. condition="PCM"
  162. println("testing: ",condition)
  163. flush(stdout)
  164. saveat=1/44100
  165. response_sr=Int64(1/saveat)
  166. C=p.C
  167. vsyn=p.vsyn
  168. frequencies=range(freq_range[1],freq_range[2],length=60)#[1:50] #cut to 50 to avoid failed simulations at high freq end.
  169. @info "Frequencies being tested: $(frequencies)"
  171. all_60_trials=Vector{Any}()
  172. all_stimuli=Vector{Any}()
  173. for stimulus_idx in 1:3 #3 sets of 20 stimuli to cover 60 frequencies, structured this way to match prior phoneme drive generation.
  174. println("stimulus set: ",stimulus_idx)
  175. p.noise_case_reference=stimulus_idx
  176. p.noise_selector=1
  177. flush(stdout)
  178. #set stimulus based on type: either normal envelopes, or derivative or event based impulse trains (peak rate or peak envelope)
  179. if stimulus_type=="envelope"
  180. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  182. elseif stimulus_type=="derivative"
  183. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  184. interpolators_global=get_smooth_derivative(interpolators_global,Float64(response_sr))
  185. elseif stimulus_type=="peakrate"
  186. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  187. interpolators_global=get_scaled_peakrate_impulses(interpolators_global,Float64(response_sr))
  188. elseif stimulus_type=="peakenvelope"
  189. global interpolators_global = jldopen("./Sine_Drives/drive_interpolators_$(stimulus_idx).jld2","r")["drives"]
  190. interpolators_global=get_unitary_peakenv_impulses(interpolators_global,Float64(response_sr))
  191. else
  192. error("Unknown stimulus type: $(stimulus_type), must be one of 'envelope', 'derivative', 'peakrate', or 'peakenvelope'.")
  193. end
  194. println("running sim")
  195. flush(stdout)
  196. if stimulus_idx==3 #to match the cut to 50 frequencies above. #no longer cut.
  197. trialdata=Ensemble_NoisyPhoneme_PCM(vary_noise_and_initial_conditions_NGNMM,time_range,p,u_init,20,saveat)
  198. else
  199. trialdata=Ensemble_NoisyPhoneme_PCM(vary_noise_and_initial_conditions_NGNMM,time_range,p,u_init,20,saveat)
  200. end
  201. push!(all_60_trials,trialdata...)
  202. push!(all_stimuli,interpolators_global...)
  203. end
  205. PCM_vectors,rates=calculate_PCM_NGNMM_noisyrates(all_60_trials,all_stimuli,frequencies,response_sr,time_range,C,vsyn,PCMrange) #ITPCrange can now exclude first 500ms of stimulus as per O.Cucu paper.
  206. #save the rates for each trial, for comparison across natural frequencies, and computation of congolmerate activity if desired.
  207. if save_traj
  208. #check dir exists:
  209. if !isdir(savepath*"Trajectories/")
  210. mkdir(savepath*"Trajectories/")
  211. end
  212. local name="Trajectories/fast_NGNMM_response_to_abs_sine_stimulus_freq_range_$(freq_range[1])to$(freq_range[2])Hz_$(stimulus_type)_transformed"
  213. CSV.write(savepath*name*".csv",DataFrame(rates,:auto),writeheader=true)
  214. generate_arrow(name,savepath)
  215. rm(savepath*name*".csv")
  216. end
  217. return(PCM_vectors)
  218. end
  221. function gaussian_filter_and_hilbert_for_PCM(signal::Vector{Float64},sampling_rate,peak_frequency::Float64)
  222. num_samples=length(signal)
  223. #put in frequency domain
  224. freqs=fftfreq(num_samples,sampling_rate)
  225. signal_fft=fft(signal)
  227. #gaussian filter parameters
  228. center_frequency=peak_frequency
  229. filter_width=center_frequency/2.0
  231. #gaussian filter:
  232. gauss_kernel=exp.(-0.5.*((abs.(freqs).-center_frequency)./(filter_width)).^2)
  233. #apply in freq domain;
  234. filtered_signal_fft=signal_fft.*gauss_kernel
  235. #return to time domain:
  236. filtered_signal=ifft(filtered_signal_fft)
  237. #return envelope via hilbert
  238. analytic_signal=hilbert(real.(filtered_signal))
  239. return analytic_signal
  240. end
  242. function calculate_PCM_oscillator_noisyrates(vector_of_solutions,stimuli,frequencies,response_sampling_rate,timerange,PCMrange,noise_seed=123)
  243. num_trials=length(vector_of_solutions)
  245. rates_sr=response_sampling_rate
  246. #add noise to the model response ('firing rates'):
  247. rates=Array{Vector}(undef,num_trials)
  248. for (idx,data) in enumerate(vector_of_solutions)
  249. rates[idx]= coupled_oscillator_activity(data)[Int64(PCMrange[1]*rates_sr):Int64(PCMrange[2]*rates_sr)]#cos(theta).*r for coupled oscillator activity
  250. trialwise_seed=noise_seed+idx
  251. noises=seeded_noise(trialwise_seed, 1.0, 0.0, length(rates[idx])) #1/f noise with specific seed, different over the 60 trials, but the same over external parameter sets.
  252. noise_power=mean(noises.^2)
  253. rate_power=mean(rates[idx].^2)
  254. # desired_signal_to_noise_ratio=0.5 #for high SNR test.
  255. desired_signal_to_noise_ratio=100.0 #for clean test.
  256. noise_scaling_factor=sqrt(rate_power/(noise_power*desired_signal_to_noise_ratio))
  257. scaled_noise=noises.*noise_scaling_factor
  258. noisy_rates=(rates[idx]).+(scaled_noise)
  259. rates[idx]=noisy_rates
  260. end
  261. #compute PCM vectors.
  262. mean_phase_diff_vectors=Vector{ComplexF64}(undef,num_trials)
  263. for trial_idx in 1:num_trials
  264. #get the stimulus for this trial:
  265. stimulus_interpolator=stimuli[trial_idx]
  266. #get the peak frequency of this stimulus:
  267. peak_frequency=frequencies[trial_idx]
  268. #turn stimulus interpolator into values over time range matching the response sampling rate:
  269. # response_idx=range(PCMrange[1]*44100,PCMrange[2]*44100,step=1)
  270. response_idx=range(PCMrange[1]*rates_sr,PCMrange[2]*rates_sr,step=1)
  271. # response_times=response_idx./44100.0
  272. response_times=response_idx./rates_sr
  273. # stimulus_idxs=response_times.*44100.0 #stimulus at 44100Hz sampling rate. (hang over from phoneme drive interpolator.)
  274. stimulus_idxs=response_times.*rates_sr #stimulus at 44100Hz sampling rate. (hang over from phoneme drive interpolator.)
  275. stimulus_values=1.0 .+ stimulus_interpolator(stimulus_idxs)
  276. #filter and get instantaneous phase:
  277. if trial_idx==1
  278. @info size(rates[trial_idx])
  279. @info size(stimulus_values)
  280. end
  281. filtered_response=gaussian_filter_and_hilbert_for_PCM(rates[trial_idx],rates_sr,peak_frequency)
  282. filtered_stimulus=gaussian_filter_and_hilbert_for_PCM(stimulus_values,rates_sr,peak_frequency)
  283. #get phases
  284. instantaneous_response_phase=angle.(filtered_response)
  285. instantaneous_stimulus_phase=angle.(filtered_stimulus)
  286. #get phase diff:
  287. phase_diff=instantaneous_response_phase.-instantaneous_stimulus_phase
  288. #convert to complex unit vector form:
  289. phase_diff_vectors=exp.(im.*phase_diff)
  290. #get average phase difference over time
  291. mean_phase_diff_vectors[trial_idx]=mean(phase_diff_vectors)
  292. end
  293. PCM_vectors=mean_phase_diff_vectors
  294. return PCM_vectors,rates
  295. end
  297. function calculate_PCM_EvokedModel_noisyrates(vector_of_solutions,stimuli,frequencies,response_sampling_rate,timerange,PCMrange,noise_seed=123)
  298. num_trials=length(vector_of_solutions)
  300. rates_sr=response_sampling_rate
  301. #add noise to the model response ('firing rates'):
  302. rates=Array{Vector}(undef,num_trials)
  303. for (idx,data) in enumerate(vector_of_solutions)
  304. rates[idx]= data[1,Int64(PCMrange[1]*rates_sr):Int64(PCMrange[2]*rates_sr)] #just the first component of the solution, the 'activity' of the evoked model, similar to global conductance in NGNMM.
  305. trialwise_seed=noise_seed+idx
  306. noises=seeded_noise(trialwise_seed, 1.0, 0.0, length(rates[idx])) #1/f noise with specific seed, different over the 60 trials, but the same over external parameter sets.
  307. noise_power=mean(noises.^2)
  308. rate_power=mean(rates[idx].^2)
  309. # desired_signal_to_noise_ratio=0.5 #for high SNR test.
  310. desired_signal_to_noise_ratio=100.0 #for clean test.
  311. noise_scaling_factor=sqrt(rate_power/(noise_power*desired_signal_to_noise_ratio))
  312. scaled_noise=noises.*noise_scaling_factor
  313. noisy_rates=(rates[idx]).+(scaled_noise)
  314. rates[idx]=noisy_rates
  315. end
  316. #compute PCM vectors.
  317. mean_phase_diff_vectors=Vector{ComplexF64}(undef,num_trials)
  318. for trial_idx in 1:num_trials
  319. #get the stimulus for this trial:
  320. stimulus_interpolator=stimuli[trial_idx]
  321. #get the peak frequency of this stimulus:
  322. peak_frequency=frequencies[trial_idx]
  323. #turn stimulus interpolator into values over time range matching the response sampling rate:
  324. response_idx=range(PCMrange[1]*rates_sr,PCMrange[2]*rates_sr,step=1)
  325. response_times=response_idx./rates_sr
  326. stimulus_idxs=response_times.*rates_sr #stimulus at 44100Hz sampling rate. (hang over from phoneme drive interpolator.)
  327. stimulus_values=stimulus_interpolator(stimulus_idxs)
  328. #filter and get instantaneous phase:
  329. if trial_idx==1
  330. @info size(rates[trial_idx])
  331. @info size(stimulus_values)
  332. end
  333. filtered_response=gaussian_filter_and_hilbert_for_PCM(rates[trial_idx],rates_sr,peak_frequency)
  334. filtered_stimulus=gaussian_filter_and_hilbert_for_PCM(stimulus_values,rates_sr,peak_frequency)
  335. #get phases
  336. instantaneous_response_phase=angle.(filtered_response)
  337. instantaneous_stimulus_phase=angle.(filtered_stimulus)
  338. #get phase diff:
  339. phase_diff=instantaneous_response_phase.-instantaneous_stimulus_phase
  340. #convert to complex unit vector form:
  341. phase_diff_vectors=exp.(im.*phase_diff)
  342. #get average phase difference over time
  343. mean_phase_diff_vectors[trial_idx]=mean(phase_diff_vectors)
  344. end
  345. PCM_vectors=mean_phase_diff_vectors
  346. return PCM_vectors,rates
  347. end
  349. function calculate_PCM_NGNMM_noisyrates(vector_of_solutions,stimuli,frequencies,response_sampling_rate,timerange,C,vsyn,PCMrange,noise_seed=123)
  350. num_trials=length(vector_of_solutions)
  352. rates_sr=response_sampling_rate
  353. #add noise to the model response ('firing rates'):
  354. rates=Array{Vector}(undef,num_trials)
  355. for (idx,data) in enumerate(vector_of_solutions)
  356. rates[idx]=get_firing_rate_NMM(data,C,vsyn)[1][Int64(PCMrange[1]*rates_sr):Int64(PCMrange[2]*rates_sr)]#
  357. trialwise_seed=noise_seed+idx
  358. noises=seeded_noise(trialwise_seed, 1.0, 0.0, length(rates[idx])) #1/f noise with specific seed, different over the 60 trials, but the same over external parameter sets.
  359. noise_power=mean(noises.^2)
  360. rate_power=mean(rates[idx].^2)
  361. # desired_signal_to_noise_ratio=0.5 #for high SNR test.
  362. desired_signal_to_noise_ratio=100.0 #for clean test.
  363. noise_scaling_factor=sqrt(rate_power/(noise_power*desired_signal_to_noise_ratio))
  364. scaled_noise=noises.*noise_scaling_factor
  365. noisy_rates=(rates[idx]).+(scaled_noise)
  366. rates[idx]=noisy_rates
  367. end
  368. #compute PCM vectors.
  369. mean_phase_diff_vectors=Vector{ComplexF64}(undef,num_trials)
  370. for trial_idx in 1:num_trials
  371. #get the stimulus for this trial:
  372. stimulus_interpolator=stimuli[trial_idx]
  373. #get the peak frequency of this stimulus:
  374. peak_frequency=frequencies[trial_idx]
  375. if trial_idx==1
  376. @info "peak frequency for trial 1: $(peak_frequency)"
  377. end
  378. #turn stimulus interpolator into values over time range matching the response sampling rate:
  379. response_idx=range(PCMrange[1]*rates_sr,PCMrange[2]*rates_sr,step=1)
  380. response_times=response_idx./rates_sr
  381. stimulus_idxs=response_times.*rates_sr #stimulus at 44100Hz sampling rate. (hang over from phoneme drive interpolator.)
  382. # stimulus_values=stimulus_interpolator(stimulus_idxs)
  383. stimulus_values=1.0 .+(stimulus_interpolator(stimulus_idxs)) #set to abs to match what the NGNMM recieved.
  384. #filter and get instantaneous phase:
  385. if trial_idx==1
  386. @info size(rates[trial_idx])
  387. @info size(stimulus_values)
  388. end
  389. filtered_response=gaussian_filter_and_hilbert_for_PCM(rates[trial_idx],rates_sr,peak_frequency)
  390. filtered_stimulus=gaussian_filter_and_hilbert_for_PCM(stimulus_values,rates_sr,peak_frequency)
  391. #get phases
  392. instantaneous_response_phase=angle.(filtered_response)
  393. instantaneous_stimulus_phase=angle.(filtered_stimulus)
  394. #get phase diff:
  395. phase_diff=instantaneous_response_phase.-instantaneous_stimulus_phase
  396. #convert to complex unit vector form:
  397. phase_diff_vectors=exp.(im.*phase_diff)
  398. #get average phase difference over time
  399. mean_phase_diff_vectors[trial_idx]=mean(phase_diff_vectors)
  400. end
  401. PCM_vectors=mean_phase_diff_vectors
  402. return PCM_vectors,rates
  403. end

Phase_Concentration_Metric.jl at commit 7df5e28, under MIT · at the source

Overview

Authors: Andrew Shannon1, David Andrew William Barton2, Martin Homer2, Conor Houghton2
ORCID iDs: Andrew Shannon, David Andrew William Barton, Martin Homer, Conor Houghton
  1. School of Computer Science, University of Bristol, Bristol BS8 1TH, United Kingdom
  2. School of Engineering Mathematics and Technology, University of Bristol, Bristol BS8 1TH, United Kingdom
Institutions: University of Bristol (United Kingdom)
Journal: eNeuro, volume 13, issue 8, pages ENEURO.0379-25.2026
Dates: received 14 October 2025; accepted 24 June 2026; published online 21 August 2026; in print August 2026
Type: Research article · Language: English
License: CC BY
Identifiers: DOI 10.1523/eneuro.0379-25.2026 · PMID 42547453 · PMCID PMC13505880 · OpenAlex W4415372432
Open access: gold, a free copy (OpenAlex)
Preprint: doi:10.1101/2025.10.20.683434
Status: code verified
Categories: EEG (modality), human (organism)
Methods: Spectral & time-frequency, Statistics, Connectivity, Preprocessing, Single-unit activity, calcium imaging
Keywords: biophysical modeling, mathematical modeling, neural mass model, speech processing, syllable segregation
MeSH: Models, Neurological*, Speech Perception*, Acoustic Stimulation, Computer Simulation, Electroencephalography, Humans (* major topic)
Journal subjects: Research Article: New Research, Cognition and Behavior
Topic: Neural Networks and Applications (Artificial Intelligence, Computer Science), according to OpenAlex
Funding: URKI | EPSRC | Centre for Doctoral Training in Interactive Artificial Intelligence (EP/S022937/1)
Citations: not cited yet (Europe PMC); 29 references in the paper

Abstract

Segregation of speech into syllables is a key step in neural speech processing. It relies on the alignment of neural activity with the rhythmic structure of speech. Two competing hypotheses explain this “neural speech tracking”, phase-resetting and evoked responses. While phenomenological modeling of these hypotheses has been successful, we still lack understanding of the underlying cortical circuits. To investigate these mechanisms, we evaluate whether a biophysical next-generation neural mass model (NMM) can reproduce several features of neural speech tracking, using phenomenological models of the competing hypotheses as algorithmic baselines. We investigate the models’ dynamics with four tests: recreating in silico an EEG experiment that identified a correlation between tracking strength and phoneme sharpness, computing the phase concentration metric, testing the effect of varying syllabic rates, and evaluating the inter event phase coherence (IEPC) across phoneme onsets. While all of the models that we study reproduce the sharpness-tuned rhythmic speech tracking, the evoked model requires a pre-processed acoustic edge impulse stimulus. We demonstrate that the NMM is performing thresholded phase-resetting triggered by sharp onsets in the continuous speech envelope. This produces cross-frequency nested oscillations that qualitatively match an experimentally-observed dual-peak signature in the IEPC. Our results indicate that the biophysical NMM provides a mechanistic bridge between generic oscillatory dynamics in cortical populations and the cognitive computations of speech tracking. Indeed, the nonlinear dynamics of the NMM offer an explanation for how peak-rate event representations in auditory cortex activity arise in response to continuous acoustic input.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 21 matches between paragraphs and lines of code.

Modelling-Speech-Processing/Shannon_et_al

License: MIT
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Commit: 7df5e283a78c36c77fc39686b403526c03b4c553, 25 June 2026
Languages: Julia (19)
Size: 377 files, 19 scripts
Software Heritage: not archived
Found in: “Code accessibility”
Holds: README, license file, environment (Manifest.toml, Project.toml)
Not found: CITATION.cff, tests, continuous integration, documentation
Tools: Plots.jl (18 files), DataFrames.jl (7 files), DifferentialEquations.jl (6 files), Distributions.jl (2 files)
Availability: 1 check, the latest on 27 September 2026: the link answers
  • 27 September 2026: the link answers
21 files
At the source: github.com/Modelling-Speech-Processing/Shannon_et_al

Code accessibility

The code/software and data described in the paper is freely available online at https://github.com/Modelling-Speech-Processing/Shannon_et_al, and the code is also provided in Extended Data 1 (https://doi.org/10.1523/ENEURO.0379-25.2026.d1). In addition, the code and data are available at the University of Bristol data repository, data.bris, at https://doi.org/10.5523/bris.2796nqydg7fub27scai9muahlr.

Reproduced under the paper's license (CC BY), from the paper cited above.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 19 scripts, each with its path and the digest of its content;
  • 21 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 4 authors, 5 keywords, 6 MeSH terms, 1 funder, 28 references.

Cite

This paper

Shannon, A., Barton, D. A. W., Homer, M., & Houghton, C. (2026). Next-Generation Neural Mass Models Reproduce Features of Speech Processing. eNeuro, 13(8), ENEURO.0379-25.2026. https://doi.org/10.1523/eneuro.0379-25.2026

BibTeX

@article{shannon2026next,
author = {Shannon, Andrew and Barton, David Andrew William and Homer, Martin and Houghton, Conor},
title = {{Next-Generation Neural Mass Models Reproduce Features of Speech Processing}},
journal = {eNeuro},
year = {2026},
month = aug,
volume = {13},
number = {8},
pages = {ENEURO.0379--25.2026},
publisher = {Society for Neuroscience},
issn = {2373-2822},
doi = {10.1523/eneuro.0379-25.2026},
url = {https://doi.org/10.1523/eneuro.0379-25.2026},
pmid = {42547453},
pmcid = {PMC13505880}
}

RIS

TY - JOUR
AU - Shannon, Andrew
AU - Barton, David Andrew William
AU - Homer, Martin
AU - Houghton, Conor
TI - Next-Generation Neural Mass Models Reproduce Features of Speech Processing
T2 - eNeuro
J2 - eNeuro
PY - 2026
DA - 2026/08/24
VL - 13
IS - 8
SP - ENEURO.0379
EP - 25.2026
SN - 2373-2822
PB - Society for Neuroscience
DO - 10.1523/eneuro.0379-25.2026
UR - https://doi.org/10.1523/eneuro.0379-25.2026
LA - en
ER -

CSL-JSON

{
"id": "10.1523/eneuro.0379-25.2026",
"type": "article-journal",
"title": "Next-Generation Neural Mass Models Reproduce Features of Speech Processing",
"container-title": "eNeuro",
"author": [
{
"family": "Shannon",
"given": "Andrew"
},
{
"family": "Barton",
"given": "David Andrew William"
},
{
"family": "Homer",
"given": "Martin"
},
{
"family": "Houghton",
"given": "Conor"
}
],
"container-title-short": "eNeuro",
"volume": "13",
"issue": "8",
"page": "ENEURO.0379-25.2026",
"DOI": "10.1523/eneuro.0379-25.2026",
"PMID": "42547453",
"PMCID": "PMC13505880",
"ISSN": "2373-2822",
"publisher": "Society for Neuroscience",
"URL": "https://doi.org/10.1523/eneuro.0379-25.2026",
"language": "en",
"issued": {
"date-parts": [
[
2026,
8,
24
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1371/journal.pcbi.1014222 [code]
Neural population models for EEG: From Canonical models to alternative model structures.
Journal: PLoS computational biology
In common: DifferentialEquations.jl, DataFrames.jl, Plots.jl, 1 other tool, EEG, 2 references
[2] doi:10.1038/s42003-026-10957-8 [code]
Brain defence by the extracellular matrix protein Cochlin.
Journal: Communications biology
In common: DifferentialEquations.jl, DataFrames.jl, Plots.jl, 1 other tool
[3] doi:10.1038/s41540-026-00749-5 [code]
A novel approach to quantify out-of-distribution uncertainty in Neural and Universal Differential Equations.
Journal: NPJ systems biology and applications
In common: DifferentialEquations.jl, DataFrames.jl, Plots.jl, 1 other tool
[4] doi:10.1371/journal.pcbi.1014549 [code]
Inward rectifier potassium channels interact with calcium channels to promote robust and physiological bistability.
Journal: PLoS computational biology
In common: DifferentialEquations.jl, DataFrames.jl, Plots.jl, 1 other tool
[5] doi:10.1093/pnasnexus/pgag213 [code]
Two-factor synaptic plasticity enables memory consolidation during neuronal burst firing.
Journal: PNAS nexus
In common: DifferentialEquations.jl, DataFrames.jl, Plots.jl, 1 other tool
[6] doi:10.1371/journal.pcbi.1014001 [code]
Burst firing creates an attractor in synaptic weight dynamics.
Journal: PLoS computational biology
In common: DifferentialEquations.jl, DataFrames.jl, Plots.jl, 1 other tool
[7] doi:10.1162/imag.a.1249 [code]
Global search metaheuristics for neural mass model calibration.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: DifferentialEquations.jl, Plots.jl, Distributions.jl, EEG
[8] doi:10.1038/s41586-026-10629-x [code]
Whole-genome duplication shaped cell-type evolution in the vertebrate brain.
Journal: Nature
In common: DataFrames.jl, Plots.jl, Distributions.jl
[9] doi:10.7554/elife.89629 [code]
Active dendrites enable robust spiking computations despite timing jitter.
Journal: eLife
In common: DifferentialEquations.jl, Plots.jl, Distributions.jl
[10] doi:10.1371/journal.pbio.3003924
Endogenous auditory and motor brain rhythms predict individual speech tracking.
Journal: PLoS biology
In common: 5 references

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.