Detecting Gender, Emotions, and Spoofing from Voice May 2017 – February 2018Stevens & Accenture Research Project· Objective: To detect gender, emotions, and spoofing attacks from voice signals with machine learningand deep neural networks.· We implemented machine learning systems and convolution deep neural networks to detect gender,emotions, and spoofing attacks with voice features such as spectrograms, mel-frequency cepstral coefficients(MFCCs), tempo-grams, chroma, and tonnetz. The detection of seven emotions reached 99%accuracy when evaluated with SAVEE dataset. We propose the use of speed features in addition to thetraditional features to detect various spoofing models. The spoofing attacks were detected with about94% accuracy when evaluated with Automatic Speaker Verification Spoofing and CountermeasuresChallenge (ASVspoof 2015) Database.· We developed a Python package for voice spoofing detection and it turned into a US patent.