Source code for bci_framework.extensions.data_analysis.utils

"""
=====
Utils
=====

This module define usefull decorators to use with data analysis.
"""

import os
import json
import time
import logging
import random
from datetime import datetime, timedelta
from multiprocessing import Process
from threading import Thread
from typing import Callable
import re

import numpy as np
from openbci_stream.acquisition import OpenBCIConsumer

from ...extensions import properties as prop


class data:
    value = {
        'context': {},
    }


data_tmp_aux_ = None
data_tmp_eeg_ = None


# ----------------------------------------------------------------------
[docs]def subprocess_this(fn: Callable) -> Callable: """Decorator to move methods to subprocessing.""" def wraper(*args, **kwargs): c = Process(target=fn, args=args) c.start() return wraper
# ----------------------------------------------------------------------
[docs]def thread_this(fn: Callable) -> Callable: """Decorator to move methods to threading.""" def wraper(*args, **kwargs): c = Thread(target=fn, args=args) c.start() return wraper
# ----------------------------------------------------------------------
[docs]def timeit(fn: Callable) -> Callable: """Decorator to calculate the execution time of a method.""" def wraper(self, *args, **kwargs): t0 = time.time() r = fn(self, *args, **kwargs) t1 = time.time() print(f"[timeit] {fn.__name__}: {(t1-t0)*1000:.2f} ms") return r return wraper
# ----------------------------------------------------------------------
[docs]def loop_consumer(*topics, package_size=None) -> Callable: """Decorator to iterate methods with new streamming data. This decorator will call a method on every new data streamming input. """ global data_tmp_eeg_, data_tmp_aux_, package_size_ topics = list(topics) if json.loads(os.getenv('BCISTREAM_RASPAD')): package_size_ = 1000 else: package_size_ = package_size def wrap_wrap(fn: Callable) -> Callable: global data_tmp_eeg_, data_tmp_aux_, package_size_ arguments = fn.__code__.co_varnames[1:fn.__code__.co_argcount] def wrap(cls): global data_tmp_eeg_, data_tmp_aux_, package_size_ if cls._feedback: topics.append('feedback') # if cls._package_size: # package_size_ = cls._package_size with OpenBCIConsumer(host=prop.HOST, topics=topics) as stream: frame = 0 for data in stream: if cls._package_size: package_size_ = cls._package_size if data.topic == 'feedback': feedback = data.value if feedback['name'] == cls._feedback.name and feedback['mode'] == 'stimuli2analysis': cls._feedback._on_feedback(**feedback) continue if data.topic == 'eeg': frame += 1 if hasattr(cls, 'buffer_eeg_'): cls.update_buffer( eeg=data.value['data'], timestamp=min(data.value['context']['timestamp.binary']) - prop.OFFSET) data_ = data.value['data'] elif data.topic == 'aux': frame += 1 if hasattr(cls, 'buffer_aux_'): cls.update_buffer( aux=data.value['data'], timestamp=min(data.value['context']['timestamp.binary']) - prop.OFFSET) data_ = data.value['data'] else: data_ = data.value # latency calculated with `timestamp.binary` if data.topic in ['eeg', 'aux']: latency = (datetime.now() - datetime.fromtimestamp( min(data.value['context']['timestamp.binary']) - prop.OFFSET)).total_seconds() * 1000 else: # latency calculated with kafka timestamp latency = (datetime.now( ) - datetime.fromtimestamp(data.timestamp / 1000)).total_seconds() * 1000 if package_size_ and (data.topic in ['eeg', 'aux']): if data.topic == 'eeg': if data_tmp_eeg_ is None: data_tmp_eeg_ = np.zeros( (data_.shape[0], 0)) data_tmp_eeg_ = np.concatenate( [data_tmp_eeg_, data_], axis=1) d = data_tmp_eeg_ elif data.topic == 'aux': if data_tmp_aux_ is None: data_tmp_aux_ = np.zeros((data_.shape[0], 0)) data_tmp_aux_ = np.concatenate( [data_tmp_aux_, data_], axis=1) d = data_tmp_aux_ kwargs = {'data': d, 'kafka_stream': data, 'topic': data.topic, 'frame': frame, 'latency': latency, } n = (package_size_ // prop.STREAMING_PACKAGE_SIZE) if frame % n == 0: fn(*[cls] + [kwargs[v] for v in arguments]) # else: if data.topic == 'eeg': data_tmp_eeg_ = np.zeros((data_.shape[0], 0)) elif data.topic == 'aux': data_tmp_aux_ = np.zeros((data_.shape[0], 0)) else: kwargs = {'data': data_, 'kafka_stream': data, 'topic': data.topic, 'frame': frame, 'latency': latency, } fn(*[cls] + [kwargs[v] for v in arguments]) return wrap return wrap_wrap
# ----------------------------------------------------------------------
[docs]def fake_loop_consumer(*topics, package_size=None) -> Callable: """Decorator to iterate methods with new streamming data. This decorator will call a method with fake data. """ # ---------------------------------------------------------------------- def wrap_wrap(fn: Callable) -> Callable: arguments = fn.__code__.co_varnames[1:fn.__code__.co_argcount] def wrap(cls): frame = 0 while True: frame += 1 t0 = time.time() num_data = int(prop.STREAMING_PACKAGE_SIZE) num_data = random.randint(num_data - 10, num_data + 10) eeg = 150 * np.random.normal(0, 0.2, size=( len(prop.CHANNELS), num_data)) if prop.BOARDMODE == 'default': aux = np.random.normal(0, 0.2, size=(3, num_data)) elif prop.BOARDMODE == 'analog': if prop.CONNECTION == 'wifi': aux = np.random.normal(0, 0.07, size=(3, num_data)) if (frame // 10) % 2: aux += 100 else: aux = np.random.normal(0, 0.07, size=(3, num_data)) if (time.time() // 1) % 2: aux += 1 elif prop.BOARDMODE == 'digital': if prop.CONNECTION == 'wifi': aux = np.random.normal(0, 0.2, size=(3, num_data)) else: aux = np.random.normal(0, 0.2, size=(5, num_data)) else: aux = None data.timestamp = datetime.now().timestamp() * 1000 data.value['timestamp'] = datetime.now() # data.value['data'] = eeg, aux if 'eeg' in topics: if hasattr(cls, 'buffer_eeg'): cls.update_buffer( eeg=eeg, timestamp=data.value['timestamp'].timestamp()) kwargs = {'data': eeg, 'kafka_stream': data, 'topic': 'eeg', 'frame': frame, 'latency': 0, } fn(*[cls] + [kwargs[v] for v in arguments]) if 'aux' in topics: if hasattr(cls, 'buffer_aux'): cls.update_buffer( aux=aux, timestamp=data.value['timestamp'].timestamp()) kwargs = {'data': aux, 'kafka_stream': data, 'topic': 'eeg', 'frame': frame, 'latency': 0, } fn(*[cls] + [kwargs[v] for v in arguments]) if 'marker' in topics: if np.random.random() > 0.9: data.value['timestamp'] = datetime.now() # data.value['datetime'] = datetime.now() data.value['context']['timestamp.binary'] = datetime.now() # data.value['data'] = chr( # np.random.choice(range(ord('A'), ord('Z') + 1))) # data.value['data'] = random.choice( # ['Right', 'Left', 'Up', 'Bottom']) data.value['marker'] = random.choice(['MARKER']) kwargs = {'data': data.value, 'kafka_stream': data, 'topic': 'marker', 'frame': frame, 'latency': 0, } fn(*[cls] + [kwargs[v] for v in arguments]) while time.time() < (t0 + 1 / (prop.SAMPLE_RATE / prop.STREAMING_PACKAGE_SIZE)): time.sleep(0.0001) return wrap return wrap_wrap
# ---------------------------------------------------------------------- def marker_slicing(markers, t0, t1): """""" if isinstance(markers, str): markers = [markers] def wrap_wrap(fn): arguments = fn.__code__.co_varnames[1:fn.__code__.co_argcount] def wrap(cls): cls._target_marker = [] @loop_consumer('aux', 'marker') def marker_slicing_(cls, topic, data, kafka_stream, latency): if topic == 'marker': # if data['marker'] in markers: # cls._target_marker.append( # [data['marker'], kafka_stream.value['datetime']]) if any([bool(re.match(mkr, data['marker'])) for mkr in markers]): cls._target_marker.append( [data['marker'], kafka_stream.value['datetime']]) if len(cls._target_marker) < 3: return if target := getattr(cls, '_target_marker', False): # marker, target = target last_buffer_timestamp = cls.buffer_aux_timestamp[-1] - prop.OFFSET last_target_timestamp = (datetime.fromtimestamp( target[0][1]) + timedelta(seconds=t1)).timestamp() if last_buffer_timestamp > last_target_timestamp: # if True: _marker, _target = target.pop(0) argmin = np.abs( cls.buffer_aux_timestamp - _target).argmin() start = int((prop.SAMPLE_RATE) * t0) stop = int((prop.SAMPLE_RATE) * t1) t = cls.buffer_aux_timestamp[argmin + start: argmin + stop] eeg = cls.buffer_eeg_[ :, argmin + start: argmin + stop] aux = cls.buffer_aux_[ :, argmin + start: argmin + stop] kwargs = {'eeg': eeg, 'aux': aux, 'timestamp': t, 'marker_datetime': _target, 'marker': _marker, 'latency': latency, } fn(*[cls] + [kwargs[v] for v in arguments]) else: logging.warning('Date too old to synchronize') logging.warning(f'Offset: {prop.OFFSET}') logging.warning( f'{datetime.fromtimestamp(last_buffer_timestamp), datetime.fromtimestamp(last_target_timestamp)}') marker_slicing_(cls) return wrap return wrap_wrap