Method, apparatus and computer program product for providing text independent voice conversion

Abstract

An apparatus for providing text independent voice conversion may include a first voice conversion model and a second voice conversion model. The first voice conversion model may be trained with respect to conversion of training source speech to synthetic speech corresponding to the training source speech. The second voice conversion model may be trained with respect to conversion to training target speech from synthetic speech corresponding to the training target speech. An output of the first voice conversion model may be communicated to the second voice conversion model to process source speech input into the first voice conversion model into target speech corresponding to the source speech as the output of the second voice conversion model.

What is claimed is: 
     
         1 . An apparatus comprising:
 a training device including:
 a model trainer, wherein the model trainer receives:
 a source speech to produce a training source speech, and 
 a target speech to produce a training target speech, 
 
 wherein the model trainer generates:
 a training source text, and 
 a training target text; and 
 
 a TTS device in communication with the model trainer, wherein the TTS device receives:
 the training source text to provide a training source synthetic speech, and 
 the training target text to provide a training target synthetic speech; and 
 
   a conversion device in communication with the training device, the conversion device includes:
 a first conversion model that receives the training source speech and the training source synthetic speech from the model trainer and the TTS device, respectively; and 
 a second conversion model that receives the training target speech and the training target synthetic speech from the model trainer and the TTS device, respectively, 
 wherein the first conversion model is trained using a first conversion function based on the training source speech and the training source synthetic speech and the second conversion model is trained using a second conversion function based on the training target speech and the training target synthetic speech. 
   
     
     
         2 . The apparatus of  claim 1  further comprising a processing element in communication with the training device and the conversion device, wherein the processing element controls the training device and the conversion device. 
     
     
         3 . The apparatus of  claim 1  further comprising a communication module, wherein the communication module communicates with the training device and the conversion device and the communication module provides a communication channel to a processing element remotely located from the apparatus to allow the processing element to control the training device and the conversion device through the communication module. 
     
     
         4 . The apparatus of  claim 1 , wherein the first conversion model receives the source speech and produces a synthetic speech. 
     
     
         5 . The apparatus of  claim 5 , wherein the synthetic speech is provided to the second conversion model that produces an output speech. 
     
     
         6 . The apparatus of  claim 1 , wherein an output of the first conversion model is provided to the second conversion model to produce an output speech corresponding to the source speech. 
     
     
         7 . The apparatus of  claim 1 , wherein the first conversion model further comprises a TTS module to a first synthetic source speech based on parallel text corresponding to the training source speech. 
     
     
         8 . The apparatus of  claim 1 , wherein the second conversion model further comprises a TTS module to generate the synthetic target speech based on parallel text corresponding to the training target speech. 
     
     
         9 . The apparatus of  claim 1 , wherein the first conversion model converts the source speech to an intermediate speech and the second conversion model converts the intermediate speech to an output speech. 
     
     
         10 . An apparatus comprising:
 a training device to receive:
 a source speech, and 
 a target speech, and 
   wherein the training device provides at least one of:
 a training source speech, 
 a training source text, 
 a training target speech, 
 a training target text, 
 a training source synthetic speech, and 
 a training target synthetic speech; and 
   a conversion device in communication with the training device, the conversion device includes a conversion model, the conversion device receives at least one of:
 the source speech; 
 the target speech; 
 the training source speech, 
 the training source text, 
 the training target speech, 
 the training target text, 
 the training source synthetic speech, and 
 the training target synthetic speech, 
   wherein the conversion model is trained with a first conversion function based on the training source speech and the training source synthetic speech and a second conversion function based on the training target speech and the training target synthetic speech, such that an intermediate output speech is produced based on the first conversion function applied to the source speech and an output speech is produced based on the second conversion function applied to the intermediate output speech.   
     
     
         11 . The apparatus of  claim 10  further comprising a processing element in communication with the training device and the conversion device, wherein the processing element controls the training device and the conversion device. 
     
     
         12 . The apparatus of  claim 10  further comprising a communication module, wherein the communication module communicates with the training device and the conversion device and a processing element that is remotely located from the apparatus such that the processing element controls the training device and the conversion device through the communication module. 
     
     
         13 . The apparatus of  claim 10 , wherein the training device includes:
 a model trainer that produces the training source speech, the training target speech, the training source text, and the training target text; and   a TTS device in communication with the model trainer, wherein the TTS device receives the training source text to provide a training source synthetic speech and receives the training target text to provide a training target synthetic speech.