Speech Privacy Protection based on Controlling Estimated Speech Transmission Index

Speech transmission index (STI) is an objective measurement of speech transmission quality and is used to predict the speech intelligibility. STI is also highly related with listening difficulty and is a function of room impulse response (RIR). RIR is regarded as the transfer function between the sound source and the sound receiver, and is generally unknown and non-static. This paper proposes a scheme to limit the STI to ensure speech privacy by utilizing proportional-integral- derivative (PID) control with estimated STI as the feedback value. The scheme bypasses the measuring of RIR of the environment where the conversation is being carried out, and uses a STI- estimation method and a RIR model to ensure speech privacy. The scheme manipulates the speech signal in the way that its STI has the value of 0.3 or in an unintelligible band. The STI is limited by controlling one parameter of the RIR model by using a PID controller. The performance of the scheme is evaluated through objective and subjective tests. The results indicate that the scheme is able to provide speech privacy with an average error between the actual and target STIs of 0.01.