KIAPBCI/kiap_bci/experimental/test_cereconn_rates1.py

'''
description: test spike rates from cereconn
author: Ioannis Vlachos
date:   16.11.18

based on Jonas Zimmermann test scripts

Copyright (c) 2018 Ioannis Vlachos.
All rights reserved.'''


import sys
import time

import cere_conn as cc
import numpy as np
import pylab as plt

import aux

params = aux.load_config()


# get CereConn object
ck = cc.CereConn(withSRE=True, withSBPE=False)
ck.send_open()

# Wait until connection is established
t = time.time()
while ck.get_state() != cc.ccS_Idle:
    time.sleep(0.005)
print("It took {:5.3f}s to open CereConn\n".format(time.time() - t))

# start recording
ck.send_record()


t = time.time()
while ck.get_state() != cc.ccS_Recording:
    time.sleep(0.05)
print("It took {:5.3f}s to start CereConn recording\n".format(time.time() - t))


# at least with NPlayServer, we need to wait for things to settle
time.sleep(.5)


# ck.set_spike_rate_estimator_ch_u_list([(0,1), (1,1), (0, 0), (1, 0) ])
# ck.set_spike_rate_estimator_ch_u_list([(0, 0), (1, 0), (2, 0), (3, 0), (4, 0), (5, 0)])

n_channels = 128
ck.fill_spike_rate_estimator_ch_u_list(n_channels, 1)        # record  channels, pool all units per channel

ch_map = ck.get_spike_rate_estimator_ch_u_map()

print("Channel map: {}".format(ch_map))


bin_width = params.daq.spike_rates.bin_width
run_time = int(sys.argv[1])  #sec
# run_time =  run_time  #+  bin_width / 1000 / 2.  #sec

ck.set_spike_rate_estimator_loop_interval_ms(params.daq.spike_rates.loop_interval)
# ck.set_spike_rate_estimation_method_exponential(params.daq.spike_rates.decay_factor)
ck.set_spike_rate_estimation_method_boxcar(params.daq.spike_rates.max_bins)
time.sleep(.5)

cd = ck.get_spike_rate_data()
print(cd['rates'].shape)
data = np.empty((0, n_channels))

# keep stats...
lasttime = time.time()
starttime = lasttime
times = []
samples = {}

start_ts = cd['ts']
last_ts = start_ts

print('STARTING LOOP\n')

ii = 0
while (time.time() - starttime <= run_time):
    now_time = time.time()
    print(f'Time left: {run_time - (now_time - starttime ):.1f}s.')
        
    time.sleep(min(0.5, run_time - (now_time - starttime )))
    
    cd = ck.get_spike_rate_data()
    print(f"Loop {ii}, rates shape: {cd['rates'].shape}")
    data = np.concatenate((data, cd['rates']))
    print(f'Total data shape: {data.shape}.')
    # print(np.sum(data, 1))
    
    # print("\nNSP Time since start {}s. Time since last loop {}s".format((cd['ts'] - start_ts) / 30000.0, (cd['ts'] - last_ts) / 30000.0))
    last_ts = cd['ts']

    ctime = time.time()
    times.append(ctime - lasttime)
    lasttime = ctime
    
    # print("System time since last loop: {}s".format(times[-1]))
    # print(f'time elapse: {time.time()-starttime}')
    ii += 1

    # print(data)
    # print(np.any(data))



print('\n')
print(f'Total data shape: {data.shape}')
print('\n')
print('\n')


ck.send_close()

# compute percentiles to select channel for audio feedback
pp = np.percentile(data,[5,95],axis=0)
idx = np.argsort(pp[1])
print(idx)
print(pp[0:2, idx])
plt.figure(1, figsize=(18,5))
plt.clf()
plt.bar(range(128), pp[1])
plt.bar(range(128), pp[0]+0.01)
# plt.ylim(bottom=-1)
plt.title('Firing rates percentiles')
plt.xlabel('Channel #')
plt.ylabel('sp/sec')
# plt.plot(pp.T,'-o')
plt.show()