HotWord Detection
This library provides functionality for detecting a hotword in given audio file using MFCC features and Dynamic Time Warping (DTW) pattern matching algorithm.
Installation
This project is on pypi
To install from pypi:
pip install hotword_detection
From this repository:
git clone https://github.com/sakethgsharma/HotWordDetection.git python setup.py install
Usage
Example scripts
For training a hotword, run:
python bin/trainHotword.py
For testing, run:
python bin/checkHotword.py
Supported features
- Mel Frequency Cepstral Coefficients
- Choice of selecting any suitable hot word through appropriate training paradigm
- Supports variable sampling frequencies
- Amplitude based Voice Activity Detector(VAD) used during recordings to remove extraneous noise
- Personalization using automatic DTW thresholding
MFCC Features
MFCC vectors are used in this module since they are the most commonly extracted features used for speech recognition systems.
| Parameter | Description |
|---|---|
| alpha | Parameter used in pre-emphasis filtering. Should be any value between 0 and 1. |
| N | Number of FFT points. |
| fs | Sampling frequency of stored audio file. |
| frame_dur | Duration of 1 speech frame. |
| num_filters | Number of filters used in the Mel filterbank. |
| lower_freq | Lower frequency bound used for constructing filterbank. |
| upper_freq | Upper frequency bound used for constructing filterbank. Should be less than fs/2. |
Dynamic Time Warping
Dynamic time warping (DTW) is an algorithm for measuring similarity between two temporal sequences which may vary in speed.
