Estimation of soot in real time would help in the development of engine controls during engine production to meet the emissions goals and for on-board diagnostics. This paper presents a new approach to the estimate the soot emissions from the ion current measured inside the cylinder during engine operation. The investigation was carried out on a 4.5L heavy duty, turbocharged diesel engine. The glow plug was modified to act as an ion current probe, in addition to its main function. Algorithms were developed for the ion current signal to estimate the soot formed on a cycle-by-cycle basis. A comparison was made between the estimated soot emissions and measurements made by using an opacity meter under steady state as well as under transient engine operating conditions.In this research, a non-linear multiple regression model (NLMR) was used to estimate soot percentage from the ion current signal. The ion current signal acquired in each cycle was analyzed, and a number of characterization parameters were determined then the parameters were used as inputs to the NLMR model. In addition, a sensitivity analysis for these parameters was made in order to achieve soot estimation with high accuracy. After the regression was computed, the NLMR model was tested at various engine operating conditions. The results showed a fairly good agreement between the predicted and measured values.