| Budget Amount *help |
¥1,000,000 (Direct Cost: ¥1,000,000)
Fiscal Year 1987: ¥300,000 (Direct Cost: ¥300,000)
Fiscal Year 1986: ¥700,000 (Direct Cost: ¥700,000)
|
|---|
| Research Abstract |
An adaptive digital howling canceller that continuously outputs an M-sequence and performs adaptation by using its synchronous added data is proposed and its chracteristics are examined by computer simulations. In combination with a simple voice detector which detects presence of a voice signal by comparing the signal input level of mictophone with a preset threshold value, slow changes of a howling path can be tracked and the characteristics are improved, althougth fast changes of a howling path can't be tracked because synchronous addition is used. Moreover, a preset type howling canceller that initially outputs a trainig signal, estimates the howling path and then stops its adaptation is implemented in hardware with NTT's DSP(DSSP1) and the effects of a synchronous addition are examined. It is comfirmed that adapation can be performed in the case of a low-level training signal by using synchronous addition. The hardware is built using only one DSP, therefore, the cancellable delay time of a howling path is short and it is not practical yet. To cancel howling pathes of practical, multiple DSPs or dedicated hardware will be necessary and an effective algorthm for multiprocessors will be also necessary.
Addtionally, an IIR howling canceller using a HARF(Hyper-stable Adaptive Recursive Filter) algorithm is studied to make a howling canceller which operates in a closed loop stably and to track the changes of a howling path quickly. In this case, better characteristics are comfirmed. Later on, synchronous-addition will be combined to this method to improve its characteristics.
|
