method

code/spikes_analysis.py

@@ -8,7 +8,7 @@ from IPython import embed
 # define sampling rate and data path
 sampling_rate = 40 #kHz
 data_dir = "../data"
-dataset = "2018-11-14-aa-invivo-1"
+dataset = "2018-11-14-ad-invivo-1"
 #data = ("2018-11-09-ad-invivo-1", "2018-11-09-ae-invivo-1", "2018-11-09-ag-invivo-1", "2018-11-13-aa-invivo-1",\
 #  "2018-11-13-ac-invivo-1", "2018-11-13-ad-invivo-1", "2018-11-13-ah-invivo-1", "2018-11-13-ai-invivo-1", \
 # "2018-11-13-aj-invivo-1", "2018-11-13-ak-invivo-1", "2018-11-13-al-invivo-1", "2018-11-14-aa-invivo-1", \
@@ -18,28 +18,29 @@ dataset = "2018-11-14-aa-invivo-1"
 # "2018-11-20-af-invivo-1", "2018-11-20-ag-invivo-1", "2018-11-20-ah-invivo-1", "2018-11-20-ai-invivo-1")
 
 # parameters for binning, smoothing and plotting
-#cut_window = 60
 cut_window = 20
-chirp_size = 14 #ms
+#cut_window_csi = 20 #ms
+#cut_window_plot = 50 #ms
+chirp_duration = 14 #ms
 neuronal_delay = 5 #ms
-chirp_start = int((-chirp_size/2+neuronal_delay+cut_window)*sampling_rate)
+chirp_start = int((-chirp_duration / 2 + neuronal_delay + cut_window * 2) * sampling_rate) #index
-chirp_end = int((chirp_size/2+neuronal_delay+cut_window)*sampling_rate)
+chirp_end = int((chirp_duration / 2 + neuronal_delay + cut_window * 2) * sampling_rate) #index
-num_bin = 12
+number_bins = 12
 window = 1 #ms
-time_axis = np.arange(-cut_window, cut_window, 1/sampling_rate) #steps
+time_axis = np.arange(-cut_window*2, cut_window*2, 1/sampling_rate) #steps
-spike_bins = np.arange(-cut_window, cut_window+1) #ms
+spike_bins = np.arange(-cut_window*2, cut_window*2) #ms
 
 # read data from files
 spikes = read_chirp_spikes(os.path.join(data_dir, dataset))
-eod = read_chirp_eod(os.path.join(data_dir, dataset))
+#eod = read_chirp_eod(os.path.join(data_dir, dataset))
-chirp_times = read_chirp_times(os.path.join(data_dir, dataset))
+#chirp_times = read_chirp_times(os.path.join(data_dir, dataset))
 
 # make a delta f map for the quite more complicated keys
 df_map = map_keys(spikes)
 
 # differentiate between phases
-phase_vec = np.arange(0, 1+1/num_bin, 1/num_bin)
+phase_vec = np.arange(0, 1 + 1 / number_bins, 1 / number_bins)
-cut_range = np.arange(-cut_window*sampling_rate, cut_window*sampling_rate, 1)
+cut_range = np.arange(-cut_window*2*sampling_rate, cut_window*2*sampling_rate, 1)
 
 # make dictionaries for spiketimes
 df_phase_time = {}
@@ -52,14 +53,18 @@ for deltaf in df_map.keys():
     df_phase_time[deltaf] = {}
     df_phase_binary[deltaf] = {}
     for rep in df_map[deltaf]:
+        chirp_size = int(rep[-1].strip('Hz'))
+        #print(chirp_size)
+        if chirp_size == 150:
+            continue
         for phase in spikes[rep]:
-            for idx in np.arange(num_bin):
+            for idx in np.arange(number_bins):
                 # check the phase
                 if phase[1] > phase_vec[idx] and phase[1] < phase_vec[idx+1]:
 
                     # get spikes between 50 ms before and after the chirp
                     spikes_to_cut = np.asarray(spikes[rep][phase])
-                    spikes_cut = spikes_to_cut[(spikes_to_cut > -50) & (spikes_to_cut < 50)]
+                    spikes_cut = spikes_to_cut[(spikes_to_cut > -cut_window*2) & (spikes_to_cut < cut_window*2)]
                     spikes_idx = np.round(spikes_cut*sampling_rate)
                     # also save as binary, 0 no spike, 1 spike
                     binary_spikes = np.isin(cut_range, spikes_idx)*1
@@ -80,27 +85,32 @@ csi_rates = {}
 for df in df_phase_time.keys():
     csi_trains[df] = []
     csi_rates[df] = []
-    beat_duration = int(abs(1/df*1000)) #ms
+    beat_duration = int(abs(1/df*1000)*sampling_rate) #steps
     beat_window = 0
-    while beat_window+beat_duration <= cut_window:
+    # beat window is at most 20 ms long, multiples of beat_duration
+    while beat_window+beat_duration <= cut_window*sampling_rate:
         beat_window = beat_window+beat_duration
     for phase in df_phase_time[df].keys():
 
         # csi calculation
-        # trains for synchronity and rate
+        # trains for synchrony and rate
-
         trials_binary = df_phase_binary[df][phase]
 
         train_chirp = []
         train_beat = []
-        spikerate_chirp = np.zeros(len(trials_binary))
+        #csi_spikerate = []
-        spikerate_beat = np.zeros(len(trials_binary))
         for i, trial in enumerate(trials_binary):
             smoothed_trial = smooth(trial, window, 1/sampling_rate)
             train_chirp.append(smoothed_trial[chirp_start:chirp_end])
             train_beat.append(smoothed_trial[chirp_start-beat_window:chirp_start])
-            spikerate_chirp[i] = np.mean(smoothed_trial[chirp_start:chirp_end])
+            #std_chirp = np.std(smoothed_trial[chirp_start:chirp_end])
-            spikerate_beat[i] = np.mean(smoothed_trial[chirp_start-beat_window:chirp_start])
+            #std_beat = np.std(smoothed_trial[chirp_start-beat_window:chirp_start])
+            #csi = (std_chirp - std_beat)/(std_chirp + std_beat)
+            #csi_spikerate.append(csi)
+
+        std_chirp = np.std(np.mean(train_chirp, axis=0))
+        std_beat = np.std(np.mean(train_beat, axis=0))
+        csi_spikerate = (std_chirp - std_beat) / (std_chirp + std_beat)
 
         rcs = []
         rbs = []
@@ -116,20 +126,12 @@ for df in df_phase_time.keys():
 
         r_train_chirp = np.mean(rcs)
         r_train_beat = np.mean(rbs)
-        embed()
+
-        exit()
         csi_train = (r_train_chirp - r_train_beat) / (r_train_chirp + r_train_beat)
-        csi_rate = (np.std(spikerate_chirp) - np.std(spikerate_beat)) / (np.std(spikerate_chirp) + np.std(spikerate_beat))
 
         # add the csi to the dictionaries with the correct df and phase
-        #csi_trains[df][phase] = csi_train
-        #csi_rates[df][phase] = csi_rate
-
         csi_trains[df].append(csi_train)
-        csi_rates[df].append(csi_rate)
+        csi_rates[df].append(np.mean(csi_spikerate))
-
-        #csi_trains[df].append(abs(csi_train))
-        #csi_rates[df].append(abs(csi_rate))
 
         '''
         # plot
@@ -173,3 +175,13 @@ ax.plot(np.arange(-1, len(csi_trains.keys())+1), np.zeros(len(csi_trains.keys())
 #ax.set_xticklabels(sorted(csi_trains.keys()))
 fig.tight_layout()
 plt.show()
+
+# spikerate_chirp = np.zeros(len(trials_binary))
+# spikerate_beat = np.zeros(len(trials_binary))
+# csi_trains[df][phase] = csi_train
+# csi_rates[df][phase] = csi_rate
+# csi_trains[df].append(abs(csi_train))
+# csi_rates[df].append(abs(csi_rate))
+#csi_rate = (np.std(spikerate_chirp) - np.std(spikerate_beat)) / (np.std(spikerate_chirp) + np.std(spikerate_beat))
+# spikerate_chirp[i] = np.mean(smoothed_trial[chirp_start:chirp_end])
+# spikerate_beat[i] = np.mean(smoothed_trial[chirp_start-beat_window:chirp_start])

code/stimulus_chirp.py

@@ -0,0 +1,47 @@
+import numpy as np
+import matplotlib.pyplot as plt
+
+stimulusrate = 500.  # the eod frequency of the fake fish
+currentchirptimes = [0.1]
+chirpwidth = 0.05  # ms
+chirpsize = 100.
+chirpampl = 0.02
+chirpkurtosis = 1.
+p = 0.
+stepsize = 0.00001
+
+time = np.arange(0.0, 0.2, stepsize)
+signal = np.zeros(time.shape)
+ampl = np.ones(time.shape)
+freq = np.ones(time.shape)
+
+ck = 0
+csig = 0.5 * chirpwidth / np.power(2.0*np.log(10.0), 0.5/chirpkurtosis)
+
+for k, t in enumerate(time):
+    a = 1.
+    f = stimulusrate
+    if ck < len(currentchirptimes):
+        if np.abs(t - currentchirptimes[ck]) < 2.0 * chirpwidth:
+            x = t - currentchirptimes[ck]
+            g = np.exp(-0.5 * (x/csig)**2)
+            f = chirpsize * g + stimulusrate
+            a *= 1.0 - chirpampl * g
+        elif t > currentchirptimes[ck] + 2.0 * chirpwidth:
+            ck += 1
+    freq[k] = f
+    ampl[k] = a
+    p += f * stepsize
+    signal[k] = a * np.sin(6.28318530717959 * p)
+
+fig = plt.figure()
+ax1 = fig.add_subplot(211)
+ax2 = fig.add_subplot(212)
+
+ax1.plot(time, signal)
+ax2.plot(time, freq)
+
+ax1.set_ylabel("fake fish field [rel]")
+ax2.set_xlabel("time [s]")
+ax2.set_ylabel("frequency [Hz]")
+plt.show()