diff --git a/icom/encoder.py b/icom/encoder.py
index 1887b8b..d080bbd 100644
--- a/icom/encoder.py
+++ b/icom/encoder.py
@@ -2,32 +2,36 @@
 import numpy as np
 import matplotlib.pyplot as plt
 
+import math
+
 # encode a message using poisson rate coding
 def encode_message(symbol: str,
-                   signal_rate: float,
+                   signal_freq: float,
                    duration: int,
                    **kwargs):
   codebook = kwargs.get("codebook",None)
-  noise_rate = kwargs.get("noise_rate", 0)
+  noise_freq = kwargs.get("noise_rate", 0)
 
   if codebook is None: raise TypeError("Invalid code book.")
 
 
-  rate = signal_rate + noise_rate
-
+  freq = signal_freq + noise_freq
   # generate spikes more than the required size.
-  isi = np.random.poisson(lam=rate, size= 2 * rate * duration)
+  size = math.ceil(2 * freq * duration)
+
+  isi = np.random.exponential(scale = 1/freq, size=size)
   spikes = np.cumsum(isi) # convert intervals to a train
-  spikes = spikes * (2*duration/spikes[-1]) # normalize
 
-  # add leak spikes
+  # filter out-of-bound spikes
+  spikes = spikes[np.where(spikes<duration)]
 
-  # cull spikes
-  spikes = spikes[np.where(spikes<=duration)]
+  # ! add leak spikes
+
   return spikes
 
 # generate spike train with signal rate of 2 and noise rate of 2.
-spikes = encode_message("A", 2, duration=10, codebook=['A','B'],noise_rate=2)
+spikes = encode_message("A", 4, duration=2, codebook=['A','B'],noise_rate=0)
+spikes = encode_message("A", 4, duration=2, codebook=['A','B'],noise_rate=0)
 
 
 #%% [markdown]
@@ -35,12 +39,16 @@
 # the following code plots signal train alognside the noise spikes.
 
 # generate signal spikes
-spikes = encode_message("A", 2, duration=10, codebook=['A','B'])
-
+signal_spikes = encode_message("A", 5, duration=10, codebook=['A','B'])
 # generate noise spikes
-noise_spikes = encode_message("", 2, duration=10, codebook=[])
+noise_spikes = encode_message("", 1, duration=10, codebook=[])
 
-plt.eventplot(spikes, color='white')
+spikes = np.concatenate((signal_spikes, noise_spikes))
+
+spikes = np.sort(np.unique(spikes))
+
+plt.eventplot(spikes, color = "white")
+plt.eventplot(signal_spikes, color='green')
 plt.eventplot(noise_spikes, color='red')
 plt.show()
 

diff --git a/icom/encoder.py b/icom/encoder.py
index 1887b8b..d080bbd 100644
--- a/icom/encoder.py
+++ b/icom/encoder.py
@@ -2,32 +2,36 @@
 import numpy as np
 import matplotlib.pyplot as plt
 
+import math
+
 # encode a message using poisson rate coding
 def encode_message(symbol: str,
-                   signal_rate: float,
+                   signal_freq: float,
                    duration: int,
                    **kwargs):
   codebook = kwargs.get("codebook",None)
-  noise_rate = kwargs.get("noise_rate", 0)
+  noise_freq = kwargs.get("noise_rate", 0)
 
   if codebook is None: raise TypeError("Invalid code book.")
 
 
-  rate = signal_rate + noise_rate
-
+  freq = signal_freq + noise_freq
   # generate spikes more than the required size.
-  isi = np.random.poisson(lam=rate, size= 2 * rate * duration)
+  size = math.ceil(2 * freq * duration)
+
+  isi = np.random.exponential(scale = 1/freq, size=size)
   spikes = np.cumsum(isi) # convert intervals to a train
-  spikes = spikes * (2*duration/spikes[-1]) # normalize
 
-  # add leak spikes
+  # filter out-of-bound spikes
+  spikes = spikes[np.where(spikes<duration)]
 
-  # cull spikes
-  spikes = spikes[np.where(spikes<=duration)]
+  # ! add leak spikes
+
   return spikes
 
 # generate spike train with signal rate of 2 and noise rate of 2.
-spikes = encode_message("A", 2, duration=10, codebook=['A','B'],noise_rate=2)
+spikes = encode_message("A", 4, duration=2, codebook=['A','B'],noise_rate=0)
+spikes = encode_message("A", 4, duration=2, codebook=['A','B'],noise_rate=0)
 
 
 #%% [markdown]
@@ -35,12 +39,16 @@
 # the following code plots signal train alognside the noise spikes.
 
 # generate signal spikes
-spikes = encode_message("A", 2, duration=10, codebook=['A','B'])
-
+signal_spikes = encode_message("A", 5, duration=10, codebook=['A','B'])
 # generate noise spikes
-noise_spikes = encode_message("", 2, duration=10, codebook=[])
+noise_spikes = encode_message("", 1, duration=10, codebook=[])
 
-plt.eventplot(spikes, color='white')
+spikes = np.concatenate((signal_spikes, noise_spikes))
+
+spikes = np.sort(np.unique(spikes))
+
+plt.eventplot(spikes, color = "white")
+plt.eventplot(signal_spikes, color='green')
 plt.eventplot(noise_spikes, color='red')
 plt.show()
 
diff --git a/icom/nback_structure_extractor.py b/icom/nback_structure_extractor.py
new file mode 100644
index 0000000..c5ca5b1
--- /dev/null
+++ b/icom/nback_structure_extractor.py
@@ -0,0 +1,30 @@
+# install pyreadr package to read RData file
+import pyreadr
+import pandas as pd
+
+dataset_file = './data/CL2015.rData'
+
+def count_targets(row):
+  print(row)
+  return 12
+
+# read n-back dataset
+datasets = pyreadr.read_r(dataset_file, use_objects="NB")
+dataset = datasets["NB"]
+
+# number of trials to look back for local varibilities
+window_size = 8
+
+dataset[dataset.participant=='P1'] \
+    .rolling(window_size).apply(count_targets)
+
+#dataset \
+#  .groupby(['participant','condition']) \
+
+#group by participant and condition
+#mark lures
+# targets in the block (X_t column)
+# lures in the block (X_l column)
+# X_v
+# recent targets (X_tl)
+# recetnt

diff --git a/icom/encoder.py b/icom/encoder.py
index 1887b8b..d080bbd 100644
--- a/icom/encoder.py
+++ b/icom/encoder.py
@@ -2,32 +2,36 @@
 import numpy as np
 import matplotlib.pyplot as plt
 
+import math
+
 # encode a message using poisson rate coding
 def encode_message(symbol: str,
-                   signal_rate: float,
+                   signal_freq: float,
                    duration: int,
                    **kwargs):
   codebook = kwargs.get("codebook",None)
-  noise_rate = kwargs.get("noise_rate", 0)
+  noise_freq = kwargs.get("noise_rate", 0)
 
   if codebook is None: raise TypeError("Invalid code book.")
 
 
-  rate = signal_rate + noise_rate
-
+  freq = signal_freq + noise_freq
   # generate spikes more than the required size.
-  isi = np.random.poisson(lam=rate, size= 2 * rate * duration)
+  size = math.ceil(2 * freq * duration)
+
+  isi = np.random.exponential(scale = 1/freq, size=size)
   spikes = np.cumsum(isi) # convert intervals to a train
-  spikes = spikes * (2*duration/spikes[-1]) # normalize
 
-  # add leak spikes
+  # filter out-of-bound spikes
+  spikes = spikes[np.where(spikes<duration)]
 
-  # cull spikes
-  spikes = spikes[np.where(spikes<=duration)]
+  # ! add leak spikes
+
   return spikes
 
 # generate spike train with signal rate of 2 and noise rate of 2.
-spikes = encode_message("A", 2, duration=10, codebook=['A','B'],noise_rate=2)
+spikes = encode_message("A", 4, duration=2, codebook=['A','B'],noise_rate=0)
+spikes = encode_message("A", 4, duration=2, codebook=['A','B'],noise_rate=0)
 
 
 #%% [markdown]
@@ -35,12 +39,16 @@
 # the following code plots signal train alognside the noise spikes.
 
 # generate signal spikes
-spikes = encode_message("A", 2, duration=10, codebook=['A','B'])
-
+signal_spikes = encode_message("A", 5, duration=10, codebook=['A','B'])
 # generate noise spikes
-noise_spikes = encode_message("", 2, duration=10, codebook=[])
+noise_spikes = encode_message("", 1, duration=10, codebook=[])
 
-plt.eventplot(spikes, color='white')
+spikes = np.concatenate((signal_spikes, noise_spikes))
+
+spikes = np.sort(np.unique(spikes))
+
+plt.eventplot(spikes, color = "white")
+plt.eventplot(signal_spikes, color='green')
 plt.eventplot(noise_spikes, color='red')
 plt.show()
 
diff --git a/icom/nback_structure_extractor.py b/icom/nback_structure_extractor.py
new file mode 100644
index 0000000..c5ca5b1
--- /dev/null
+++ b/icom/nback_structure_extractor.py
@@ -0,0 +1,30 @@
+# install pyreadr package to read RData file
+import pyreadr
+import pandas as pd
+
+dataset_file = './data/CL2015.rData'
+
+def count_targets(row):
+  print(row)
+  return 12
+
+# read n-back dataset
+datasets = pyreadr.read_r(dataset_file, use_objects="NB")
+dataset = datasets["NB"]
+
+# number of trials to look back for local varibilities
+window_size = 8
+
+dataset[dataset.participant=='P1'] \
+    .rolling(window_size).apply(count_targets)
+
+#dataset \
+#  .groupby(['participant','condition']) \
+
+#group by participant and condition
+#mark lures
+# targets in the block (X_t column)
+# lures in the block (X_l column)
+# X_v
+# recent targets (X_tl)
+# recetnt
diff --git a/py/poisson_generator.py b/py/poisson_generator.py
index 1bb04df..f800cfd 100644
--- a/py/poisson_generator.py
+++ b/py/poisson_generator.py
@@ -59,3 +59,18 @@
 print(np.argmax(ps), np.max(ps))
 
 #%%
+
+
+# %%
+# Using numpy exp to generate random interarrivals
+
+import numpy as np
+
+duration = 20
+rate = 5
+size = 2 * rate * duration
+
+inter_arrivals = np.random.exponential(scale = 1/rate, size=size)
+arrivals = np.cumsum(inter_arrivals)
+
+spikes = arrivals[np.where(arrivals<duration)]