10
votes
Accepted
How do AIs like Siri and Alexa respond to their names being called?
Is it because their listening function reloads in milliseconds or even nanoseconds
Yes, it expects the keyword to start every moment of time and it ignores the rest.
Overall, the algorithm is ...
- 271
4
votes
Accepted
What is a beam?
Beam search
The beam search is a algorithm to find probable output sequences for an input sequence, so it has been used for decoding in the context of sequence-to-sequence tasks, like machine ...
- 41.4k
4
votes
Term for algorithms that are not trained
The word "Artificial intelligence" refers to machines being able to have intelligence that of humans/animals. The meaning of the word was even discussed on this site. So it's up to your ...
- 216
3
votes
Open-source vocal cloning (speech-to-speech neural style transfer)
Tensorflow code for "one-to-one" style transfer:
https://github.com/phiana/speech-style-transfer-vae-gan-tensorflow
it's the implementation of a 2021 paper. Speech style transfer, voice ...
3
votes
How does the CTC loss work?
Connectionist Temporal Classification (CTC) can be useful for sequence modeling problems, like speech recognition and handwritten recognition, where the input and output sequences might have different ...
- 41.4k
2
votes
Can transformer be better than RNN for online speech recognition?
Are there examples that transformer have better accuracy than RNN end-to-end model like RNN-transducer for speech recognition?
Can transformer be used for online speech recognition which require low ...
- 271
2
votes
Term for algorithms that are not trained
How about handcrafted -as you mentioned-? In the following question it is opposed to learned.
https://datascience.stackexchange.com/questions/54390/what-is-the-difference-between-handcrafted-and-...
2
votes
Accepted
Open-source vocal cloning (speech-to-speech neural style transfer)
Additional projects that might be of interest:
Neural Voice Cloning with a Few Samples - NeurIPS 2018 (Sercan O. Arik, Jitong Chen, Kainan Peng, Wei Ping, Yanqi Zhou)
A neural voice cloning system ...
2
votes
How is speech recognition software able to distinguish between different speakers and yet still understand them all?
I would slightly disagree with Ryan's answer: the fundamental frequency is mainly specific to a speaker. Sounds are defined by other frequency patterns.
Vowels, for example, have two bands of energy ...
- 5,417
2
votes
Accepted
Has there been research done regarding processing speech then building a "speaker profile" based off the processed speech?
Yes, there is. A quick search found this:
Multimodal Speaker Identification Based on Text and Speech (2008).
In the abstract, they write
This paper proposes a novel method for speaker identification ...
- 36
2
votes
What does "use log probability to automatically increase the temperature until certain thresholds are hit" mean with OpenAI ASR with temperature=0
When temperature is set to $0$ the ASR model first selects the most likely transcription based on the logits in the most deterministic way compared to all other valid temperatures. However, when the ...
- 5,505
1
vote
How can I train AI to learn how to recognize highly impaired speech from a dystonic person?
This is typically done by supervised learning. Which begins with finding or gathering a corpus of annotated data representative of your problem. Then training an ML algorithm to minimize its loss ...
- 1,559
1
vote
What is a beam?
Here is Guillaume Klein's answer at the issue section of the Git repository:
"Beam Search" in Wikipedia:
Additionally, the beam size/beam width is controlling the number of paths that ...
- 183
1
vote
How is speech recognition software able to distinguish between different speakers and yet still understand them all?
A sound byte can be decomposed into a set of features which are distinguishable by a classifier. The most important feature extracted is the fundamental frequency. This is the lowest frequency ...
- 121
1
vote
speech comment detection by deep speech mozilla for data set
The problem you state is a well known problem, and it is called "keyword spotting" os KWS. If you add a wake up word before it (like "hey google/siri"), you can also use "...
1
vote
Studying the speech-generation model and have question about the confusing nature of model input and outputs
There are two "inputs" into Wavenet:
the previously generated samples of the waveform, which are usually encoded into multiple channels, like into 256 channels using 8-bit mu-law encoding
...
- 204
1
vote
How to use AI for language recognition?
Google has an API you can use. https://cloud.google.com/translate/. Their API can translate audio to text. They also have an API for converting speech to text. The language detection feature should ...
- 182
1
vote
What is the difference between Kaldi and DeepSpeech speech recognition systems in their approach?
Both of them using the end-to-end approach for speech recognition. However, because of the code complexity in DeepSpeech, you can't tune the model for your work. Kaldi could be configured in a ...
- 1,836
1
vote
Has there been research done regarding processing speech then building a "speaker profile" based off the processed speech?
Speaker identification is quite widely researched domain. Modern approach would be to map speaker information to i-vector, a real-valued vector of 200-400 components that characterizes speaker fully. ...
- 271
Only top scored, non community-wiki answers of a minimum length are eligible