Inverse methods in room acoustics with under-determined data and applications to virtual acoustics

With the advent of commercial surround sound systems there is a growing demand for 11 system that can convert existing mono and stereo recor~ings into a surround sound recording. A part of this problem is to generate the reverberation for the additional audio channels from the reverberation in the mono or stereo recording. . , In thi~ thesis, it is investigated whether a room model can be constructed from one or two room impulse responses and in which a virtual surround sound recording can then be perfonned. The estimation of the room model parameters is based on the three well-mown room acoustic models; the geometrical acoustic model with specular reflections, the eigenmode model and the diffuse field model. It is shown that the scope with the geometrical acoustic and the eigenmode model is limited but it is possible to obtain useful and consistent results for the room volume and the source-to-receiver distance from the diffuse field model. Based on these findings, the problem of generating multiple room impulse responses from one or two input room impulse response(s) is approached slightly differently. The very early part of the room impulse responses (the early reflections) is generated by a geometrical model with specular and diffuse reflections. The remainder of the room impulse responses are copies of the input room impulses obtained by convolution with a set of filters that control the coherence between them. The values for the coherence are given by expressions for the coherence between microphones with first-order directivity in a diffuse field and these expressions are derived in the thesis. The results from objective and subjective tests indicate that this method works successfully.EThOS - Electronic Theses Online ServiceGBUnited Kingdo