Go Down

Topic: open-source phrase database I can use to enhance speech recognition? (Read 308 times) previous topic - next topic


    So, I'm actually using python for this, and I found and downloaded the old speech module. It is technically working perfectly, but there are a lot of misreadings with speech recognition. As I understand it, most of what makes up a good speech recognition program nowadays is not necessarily the initial speech to text, but a good phrase database to compare the result to. If you've ever used Google's voice search, they made it so you can actually see in the search bar a bunch of options of what you may have said flashing by, and it picks the best one, making it one of the best if not the best speech recognition programs I've ever seen.

    Now, what I'm looking for is a database that someone has already created (or even better, a program that picks the best phrase for you!) that I can access or download, preferably for free.

    If this is not an option, do you have any other suggestions? I could just go and do it myself fairly easily, but that will take SO long.... I'm kind of keeping that as a last resort.

    In case you're interested, here's an excerpt of my code -- the part that listens for the keyword "j.j." to start listening for a command.

Code: [Select]

import speech

def listen_for_command():
#all possible misinterpretations of the keyword "j.j."
keys = ["j.j.", "g.j.", "j.g.", "g.g.", "ginger", "gigi"]
while True:
    test = speech.input().lower()

    #cycle through the possible keywords...
    trigger = False
    for key in keys:
        if key in test:
            start = test.find(key) + len(key)
            trigger = True

    #...if it was triggered, break the while loop
    if trigger:

#parse the command and return it
ret = ""
for i in range(start, len(test)):
    ret = ret + test[i]
return ret

If you only know the arduino language, I can translate if you wish. Thanks in advance!

Go Up