SpeakRite: Monitoring Speaking Rate in Real Time on a Mobile Phone

Introduction

With globalization there has been an increase in spoken interaction between people from different geographies and different cultural background. Tele conversations are increasingly replacing face to face conversations because of business economics. However, unlike face to face conversations which are essentially driven by several unspoken cues like eye contact and body language; telephone conversations are solely driven by the cues associated with speech. The main reason for ineffectiveness of tele conversations is because of the way conversation is spoken and not necessarily by what is being spoken; and the background and culture of a person determines the way the person speaks. Noticeably the ineffectiveness of tele conversation is prominent when people from different geographies or cultures converse in a language common to them. The ineffectiveness is very prominent in a call center scenario, where the call center agent and the customer are from two different geographies. Speaking rate is one critical factor affecting intelligibility and comprehension of speech (Pandharipande & Kopparapu, 2011). It is well known (Anderson-Hsieh & Koehler, 1988; Munro & Derwing, 1998; Yuan, Liberman, & Cieri, 2006; Bradlow & Pisoni, 1999) that speaking rate varies across native and non-native speakers and additionally affects comprehension of speech in native and non-native listeners. This suggests that it is important that speakers converse at an optimal speaking rate for an effective conversation. Listeners can be lost to boredom, lost to complexity or fully engaged in a conversation based on the speakers ability to deliver at the optimal rate (“Using adaptive voice”, n.d.).

For any given language the average speaking rate varies across individual speakers of that language. For instance the average English speaking rate is between 130 and 200 words per minute (WPM) (“Using adaptive voice”, n.d.). This wide WPM range applies to 90 percent of the English speaking population. It is further suggested that for speech content that is complex, a speaking rate between 130 and 145 WPM is good, while for speech content that is of average complexity, a speaking rate between 145 and 175 is optimal while for simple material, many listeners can accommodate over 175 WPM. Good communicators are aware of this and they continuously monitor their speaking rate. They consciously adjust their conversational pace to get the message across effectively and efficiently. However monitoring and maintaining the speaking rate at the desired levels, may be hard for an average person who is not conscious of the speaking rate or is in an emotional state that does not allow him to concentrate on the speaking rate. Any platform that enables automatic monitoring of speaking rate can help speakers speak at the right speed to make conversations effective. In Pandharipande and Kopparapu (2011) a server based speaking rate monitoring system that assists call center agents, in real time, to maintain an optimal speaking rate was proposed. In this paper we present SpeakRite, an application residing on an Android mobile phone, to assist a person to converse right by analyzing his spoken speech as he speaks. SpeakRite estimates the speaking rate of the conversation and provides a real time feedback to the user about his current speaking rate; this feedback can enable him to modify his speaking rate in real time. Additionally, the application provides an offline analysis of a recorded call. SpeakRite is based on the syllable count and in that sense is language independent and can be readily used for any spoken language. The main contribution of this paper is to enable a speaking rate monitoring system on a mobile phone, which can monitor the speaking rate of the mobile phone user during conversation. We discuss our computationally faster technique for monitoring speaking rate, that works on a mobile phone device in real time. We present experimental results to asses the accuracy of the proposed technique which can work in real time on a mobile device with the speaking rate system that works on the server. The rest of the paper is organized as follows: first, we discuss in brief, the server based speaking rate monitoring system (Pandharipande & Kopparapu, 2011) and then we discuss the technique to computer speaking rate on a mobile phone. We evaluate the accuracy of the proposed technique for syllable detection afterwards and we also feature of the SpeakRite Android mobile application, followed by a conclusion.