Lie Detection in speech using Machine Learning
Budget: ₹1,500 – ₹12,500 INR
The proposed method uses acoustic features in speech, such as Mel-frequency cepstral
coefficients (MFCC), energy envelopes, and pitch contours for training a recurrent neural
network to detect lies. This project aims to train a deep learning network using audio characteristics from call
recordings for lie detection. The sound characteristics that will be used for training are
pitch contours, energy envelopes and Mel-frequency cepstral coefficients (MFCC), which
are produced from a training set of deceitful and non-deceitful recordings.
Reportedly, more than 202 million spam calls were made from just one phone number
between January and October of this year. Thus, the project also aims to answer the true
intention of callers posing as bank and network operators phishing for valuable customer
data. An automated lie detection system based on acoustic features in speech can help
expose such dangerous callers and help in keeping our data secure.
coefficients (MFCC), energy envelopes, and pitch contours for training a recurrent neural
network to detect lies. This project aims to train a deep learning network using audio characteristics from call
recordings for lie detection. The sound characteristics that will be used for training are
pitch contours, energy envelopes and Mel-frequency cepstral coefficients (MFCC), which
are produced from a training set of deceitful and non-deceitful recordings.
Reportedly, more than 202 million spam calls were made from just one phone number
between January and October of this year. Thus, the project also aims to answer the true
intention of callers posing as bank and network operators phishing for valuable customer
data. An automated lie detection system based on acoustic features in speech can help
expose such dangerous callers and help in keeping our data secure.