|
Kategorie: Diplomové, bakalářské práce |
Tento dokument chci!
Práce popisuje základy principu funkčnosti neuronů a vytvoření umělých neuronových sítí. Je zde důkladně popsána struktura a funkce neuronů a ukázán nejpoužívanější algoritmus pro učení neuronů. Základy fuzzy logiky, včetně jejich výhod a nevýhod, jsou rovněž prezentovány. Detailněji je popsán algoritmus zpětného šíření chyb a adaptivní neuro-fuzzy inferenční systém. Tyto techniky poskytují efektivní způsoby učení neuronových sítí.
This network uses a
hybrid learning algorithm, effective combination neural networks and fuzzy
inference system while the other two networks are simple neural networks without the
benefits fuzzy logic reasoning.47
A slight increase efficiency (2,61%) can observed contrast NNV1 while
testing with words Speaker3 which means recognition rate words out of
153 (45,75%).
With much bigger training set containing data from various speakers the
recognition would more universal terms recognizing the isolated words by
unknown speakers.
The best results were obtained using the ANFIS network.
As was presented, both networks performed well recognizing the learned
languages especially the ones which came from the same speakers the system
was trained with. Words from
Speaker3 were allocated with almost the same accuracy with both adjustments. There was significant difference (~10÷15%) between the
efficiency recognition with the neural network depending the analysis
parameters while testing with words Speaker1 and Speaker2.
. With the frame length 200 ms, overlaps
and cepstral parameters per frame, the ANFIS network precisely identified 140
words out 153 (91,50%) while testing with words Speaker1 and 145 words out
of 153 (94,77%) while testing with words Speaker2. For Speaker3, words out of
153 were precisely identified (61,44%) which means more than 15% increase in
efficiency benchmark against the neural network. The
ANFIS network gave the best results exceeding the efficiency neural network with
more than 10÷15%
