diff --git a/mindspore/ccsrc/minddata/dataset/api/audio.cc b/mindspore/ccsrc/minddata/dataset/api/audio.cc
index aa0f33d0fdc..42a94f712f0 100644
--- a/mindspore/ccsrc/minddata/dataset/api/audio.cc
+++ b/mindspore/ccsrc/minddata/dataset/api/audio.cc
@@ -149,7 +149,7 @@ std::shared_ptr<TensorOperation> ComplexNorm::Parse() { return std::make_shared<
 
 // FrequencyMasking Transform Operation.
 struct FrequencyMasking::Data {
-  Data(bool iid_masks, int32_t frequency_mask_param, int32_t mask_start, double mask_value)
+  Data(bool iid_masks, int32_t frequency_mask_param, int32_t mask_start, float mask_value)
       : iid_masks_(iid_masks),
         frequency_mask_param_(frequency_mask_param),
         mask_start_(mask_start),
@@ -157,10 +157,10 @@ struct FrequencyMasking::Data {
   int32_t frequency_mask_param_;
   int32_t mask_start_;
   bool iid_masks_;
-  double mask_value_;
+  float mask_value_;
 };
 
-FrequencyMasking::FrequencyMasking(bool iid_masks, int32_t frequency_mask_param, int32_t mask_start, double mask_value)
+FrequencyMasking::FrequencyMasking(bool iid_masks, int32_t frequency_mask_param, int32_t mask_start, float mask_value)
     : data_(std::make_shared<Data>(iid_masks, frequency_mask_param, mask_start, mask_value)) {}
 
 std::shared_ptr<TensorOperation> FrequencyMasking::Parse() {
@@ -170,15 +170,15 @@ std::shared_ptr<TensorOperation> FrequencyMasking::Parse() {
 
 // TimeMasking Transform Operation.
 struct TimeMasking::Data {
-  Data(bool iid_masks, int64_t time_mask_param, int64_t mask_start, double mask_value)
+  Data(bool iid_masks, int32_t time_mask_param, int32_t mask_start, float mask_value)
       : iid_masks_(iid_masks), time_mask_param_(time_mask_param), mask_start_(mask_start), mask_value_(mask_value) {}
-  int64_t time_mask_param_;
-  int64_t mask_start_;
+  int32_t time_mask_param_;
+  int32_t mask_start_;
   bool iid_masks_;
-  double mask_value_;
+  float mask_value_;
 };
 
-TimeMasking::TimeMasking(bool iid_masks, int64_t time_mask_param, int64_t mask_start, double mask_value)
+TimeMasking::TimeMasking(bool iid_masks, int32_t time_mask_param, int32_t mask_start, float mask_value)
     : data_(std::make_shared<Data>(iid_masks, time_mask_param, mask_start, mask_value)) {}
 
 std::shared_ptr<TensorOperation> TimeMasking::Parse() {
diff --git a/mindspore/ccsrc/minddata/dataset/api/python/bindings/dataset/audio/kernels/ir/bindings.cc b/mindspore/ccsrc/minddata/dataset/api/python/bindings/dataset/audio/kernels/ir/bindings.cc
index 4564426ab74..af6b179ade3 100644
--- a/mindspore/ccsrc/minddata/dataset/api/python/bindings/dataset/audio/kernels/ir/bindings.cc
+++ b/mindspore/ccsrc/minddata/dataset/api/python/bindings/dataset/audio/kernels/ir/bindings.cc
@@ -134,7 +134,7 @@ PYBIND_REGISTER(
     (void)
       py::class_<audio::FrequencyMaskingOperation, TensorOperation, std::shared_ptr<audio::FrequencyMaskingOperation>>(
         *m, "FrequencyMaskingOperation")
-        .def(py::init([](bool iid_masks, int32_t frequency_mask_param, int32_t mask_start, double mask_value) {
+        .def(py::init([](bool iid_masks, int32_t frequency_mask_param, int32_t mask_start, float mask_value) {
           auto frequency_masking =
             std::make_shared<audio::FrequencyMaskingOperation>(iid_masks, frequency_mask_param, mask_start, mask_value);
           THROW_IF_ERROR(frequency_masking->ValidateParams());
@@ -146,7 +146,7 @@ PYBIND_REGISTER(
   TimeMaskingOperation, 1, ([](const py::module *m) {
     (void)py::class_<audio::TimeMaskingOperation, TensorOperation, std::shared_ptr<audio::TimeMaskingOperation>>(
       *m, "TimeMaskingOperation")
-      .def(py::init([](bool iid_masks, int64_t time_mask_param, int64_t mask_start, double mask_value) {
+      .def(py::init([](bool iid_masks, int32_t time_mask_param, int32_t mask_start, float mask_value) {
         auto time_masking =
           std::make_shared<audio::TimeMaskingOperation>(iid_masks, time_mask_param, mask_start, mask_value);
         THROW_IF_ERROR(time_masking->ValidateParams());
diff --git a/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/frequency_masking_ir.cc b/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/frequency_masking_ir.cc
index 2ee3a8a8123..a1d0b57179d 100644
--- a/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/frequency_masking_ir.cc
+++ b/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/frequency_masking_ir.cc
@@ -24,7 +24,7 @@ namespace dataset {
 namespace audio {
 
 FrequencyMaskingOperation::FrequencyMaskingOperation(bool iid_masks, int32_t frequency_mask_param, int32_t mask_start,
-                                                     double mask_value)
+                                                     float mask_value)
     : iid_masks_(iid_masks),
       frequency_mask_param_(frequency_mask_param),
       mask_start_(mask_start),
diff --git a/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/frequency_masking_ir.h b/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/frequency_masking_ir.h
index 85a23ff851f..6f710a513b5 100644
--- a/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/frequency_masking_ir.h
+++ b/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/frequency_masking_ir.h
@@ -31,7 +31,7 @@ constexpr char kFrequencyMaskingOperation[] = "FrequencyMasking";
 
 class FrequencyMaskingOperation : public TensorOperation {
  public:
-  FrequencyMaskingOperation(bool iid_masks, int32_t frequency_mask_param, int32_t mask_start, double mask_value);
+  FrequencyMaskingOperation(bool iid_masks, int32_t frequency_mask_param, int32_t mask_start, float mask_value);
 
   ~FrequencyMaskingOperation();
 
@@ -47,7 +47,7 @@ class FrequencyMaskingOperation : public TensorOperation {
   int32_t frequency_mask_param_;
   int32_t mask_start_;
   bool iid_masks_;
-  double mask_value_;
+  float mask_value_;
 };  // class FrequencyMaskingOperation
 
 }  // namespace audio
diff --git a/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/time_masking_ir.cc b/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/time_masking_ir.cc
index 0bdffac2590..bfc092db0c7 100644
--- a/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/time_masking_ir.cc
+++ b/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/time_masking_ir.cc
@@ -23,8 +23,8 @@ namespace mindspore {
 namespace dataset {
 namespace audio {
 
-TimeMaskingOperation::TimeMaskingOperation(bool iid_masks, int64_t time_mask_param, int64_t mask_start,
-                                           double mask_value)
+TimeMaskingOperation::TimeMaskingOperation(bool iid_masks, int32_t time_mask_param, int32_t mask_start,
+                                           float mask_value)
     : iid_masks_(iid_masks), time_mask_param_(time_mask_param), mask_start_(mask_start), mask_value_(mask_value) {}
 
 TimeMaskingOperation::~TimeMaskingOperation() = default;
diff --git a/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/time_masking_ir.h b/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/time_masking_ir.h
index 183ea642ed0..c8a920f3feb 100644
--- a/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/time_masking_ir.h
+++ b/mindspore/ccsrc/minddata/dataset/audio/ir/kernels/time_masking_ir.h
@@ -31,7 +31,7 @@ constexpr char kTimeMaskingOperation[] = "TimeMasking";
 
 class TimeMaskingOperation : public TensorOperation {
  public:
-  TimeMaskingOperation(bool iid_masks, int64_t time_mask_param, int64_t mask_start, double mask_value);
+  TimeMaskingOperation(bool iid_masks, int32_t time_mask_param, int32_t mask_start, float mask_value);
 
   ~TimeMaskingOperation();
 
@@ -44,10 +44,10 @@ class TimeMaskingOperation : public TensorOperation {
   Status to_json(nlohmann::json *out_json) override;
 
  private:
-  int64_t time_mask_param_;
-  int64_t mask_start_;
+  int32_t time_mask_param_;
+  int32_t mask_start_;
   bool iid_masks_;
-  double mask_value_;
+  float mask_value_;
 };  // class TimeMaskingOperation
 
 }  // namespace audio
diff --git a/mindspore/ccsrc/minddata/dataset/audio/kernels/audio_utils.cc b/mindspore/ccsrc/minddata/dataset/audio/kernels/audio_utils.cc
index d225eabd48b..45247d90f50 100644
--- a/mindspore/ccsrc/minddata/dataset/audio/kernels/audio_utils.cc
+++ b/mindspore/ccsrc/minddata/dataset/audio/kernels/audio_utils.cc
@@ -346,20 +346,20 @@ Status TimeStretch(std::shared_ptr<Tensor> input, std::shared_ptr<Tensor> *outpu
   return Status::OK();
 }
 
-Status RandomMaskAlongAxis(const std::shared_ptr<Tensor> &input, std::shared_ptr<Tensor> *output, int64_t mask_param,
-                           double mask_value, int axis, std::mt19937 rnd) {
-  std::uniform_int_distribution<int64_t> mask_width_value(0, mask_param);
+Status RandomMaskAlongAxis(const std::shared_ptr<Tensor> &input, std::shared_ptr<Tensor> *output, int32_t mask_param,
+                           float mask_value, int axis, std::mt19937 rnd) {
+  std::uniform_int_distribution<int32_t> mask_width_value(0, mask_param);
   TensorShape input_shape = input->shape();
-  int64_t mask_dim_size = axis == 1 ? input_shape[-2] : input_shape[-1];
-  int64_t mask_width = mask_width_value(rnd);
-  std::uniform_int_distribution<int64_t> min_freq_value(0, mask_dim_size - mask_width);
-  int64_t mask_start = min_freq_value(rnd);
+  int32_t mask_dim_size = axis == 1 ? input_shape[-2] : input_shape[-1];
+  int32_t mask_width = mask_width_value(rnd);
+  std::uniform_int_distribution<int32_t> min_freq_value(0, mask_dim_size - mask_width);
+  int32_t mask_start = min_freq_value(rnd);
 
   return MaskAlongAxis(input, output, mask_width, mask_start, mask_value, axis);
 }
 
-Status MaskAlongAxis(const std::shared_ptr<Tensor> &input, std::shared_ptr<Tensor> *output, int64_t mask_width,
-                     int64_t mask_start, double mask_value, int axis) {
+Status MaskAlongAxis(const std::shared_ptr<Tensor> &input, std::shared_ptr<Tensor> *output, int32_t mask_width,
+                     int32_t mask_start, float mask_value, int32_t axis) {
   if (axis != 2 && axis != 1) {
     RETURN_STATUS_UNEXPECTED("MaskAlongAxis: only support Time and Frequency masking, axis should be 1 or 2.");
   }
@@ -374,7 +374,7 @@ Status MaskAlongAxis(const std::shared_ptr<Tensor> &input, std::shared_ptr<Tenso
   CHECK_FAIL_RETURN_UNEXPECTED(mask_start + mask_width <= input_shape[check_dim_ind],
                                "MaskAlongAxis: the sum of mask_start and mask_width is out of bounds.");
 
-  int64_t cell_size = input->type().SizeInBytes();
+  int32_t cell_size = input->type().SizeInBytes();
 
   if (axis == 1) {
     // freq
diff --git a/mindspore/ccsrc/minddata/dataset/audio/kernels/audio_utils.h b/mindspore/ccsrc/minddata/dataset/audio/kernels/audio_utils.h
index 932e7e03dc2..2b8f6e21e5f 100644
--- a/mindspore/ccsrc/minddata/dataset/audio/kernels/audio_utils.h
+++ b/mindspore/ccsrc/minddata/dataset/audio/kernels/audio_utils.h
@@ -231,8 +231,8 @@ Status TimeStretch(std::shared_ptr<Tensor> input, std::shared_ptr<Tensor> *outpu
 /// \param axis: Axis to apply masking on (1 -> frequency, 2 -> time).
 /// \param rnd: Number generator.
 /// \return Status code.
-Status RandomMaskAlongAxis(const std::shared_ptr<Tensor> &input, std::shared_ptr<Tensor> *output, int64_t mask_param,
-                           double mask_value, int axis, std::mt19937 rnd);
+Status RandomMaskAlongAxis(const std::shared_ptr<Tensor> &input, std::shared_ptr<Tensor> *output, int32_t mask_param,
+                           float mask_value, int axis, std::mt19937 rnd);
 
 /// \brief Apply a mask along axis. All examples will have the same mask interval.
 /// \param input: Tensor of shape <..., freq, time>.
@@ -243,8 +243,8 @@ Status RandomMaskAlongAxis(const std::shared_ptr<Tensor> &input, std::shared_ptr
 /// \param mask_value: Value to assign to the masked columns.
 /// \param axis: Axis to apply masking on (1 -> frequency, 2 -> time).
 /// \return Status code.
-Status MaskAlongAxis(const std::shared_ptr<Tensor> &input, std::shared_ptr<Tensor> *output, int64_t mask_width,
-                     int64_t mask_start, double mask_value, int axis);
+Status MaskAlongAxis(const std::shared_ptr<Tensor> &input, std::shared_ptr<Tensor> *output, int32_t mask_width,
+                     int32_t mask_start, float mask_value, int32_t axis);
 
 /// \brief Compute the norm of complex tensor input.
 /// \param power Power of the norm description (optional).
diff --git a/mindspore/ccsrc/minddata/dataset/audio/kernels/frequency_masking_op.cc b/mindspore/ccsrc/minddata/dataset/audio/kernels/frequency_masking_op.cc
index 56bbfe09dfa..b1b5bbed06a 100644
--- a/mindspore/ccsrc/minddata/dataset/audio/kernels/frequency_masking_op.cc
+++ b/mindspore/ccsrc/minddata/dataset/audio/kernels/frequency_masking_op.cc
@@ -25,7 +25,7 @@ namespace dataset {
 
 // constructor
 FrequencyMaskingOp::FrequencyMaskingOp(bool iid_masks, int32_t frequency_mask_param, int32_t mask_start,
-                                       double mask_value)
+                                       float mask_value)
     : frequency_mask_param_(frequency_mask_param),
       mask_start_(mask_start),
       iid_masks_(iid_masks),
@@ -42,7 +42,7 @@ Status FrequencyMaskingOp::Compute(const std::shared_ptr<Tensor> &input, std::sh
   TensorShape input_shape = input->shape();
   CHECK_FAIL_RETURN_UNEXPECTED(
     input_shape[-2] >= frequency_mask_param_,
-    "FrequencyMasking: frequency_mask_param should be less than the length of frequency dimension.");
+    "FrequencyMasking: frequency_mask_param should be less than or equal to the length of frequency dimension.");
 
   std::shared_ptr<Tensor> input_tensor;
   // typecast
@@ -53,13 +53,11 @@ Status FrequencyMaskingOp::Compute(const std::shared_ptr<Tensor> &input, std::sh
   } else {
     input_tensor = input;
   }
-  auto mask_val =
-    input->type() != DataType::DE_FLOAT64 ? static_cast<float>(mask_value_) : static_cast<double>(mask_value_);
   // iid_masks - whether to apply different masks to each example/channel.
   if (iid_masks_ == false) {
-    return MaskAlongAxis(input_tensor, output, frequency_mask_param_, mask_start_, mask_val, 1);
+    return MaskAlongAxis(input_tensor, output, frequency_mask_param_, mask_start_, mask_value_, 1);
   } else {
-    return RandomMaskAlongAxis(input_tensor, output, frequency_mask_param_, mask_val, 1, rnd_);
+    return RandomMaskAlongAxis(input_tensor, output, frequency_mask_param_, mask_value_, 1, rnd_);
   }
 }
 }  // namespace dataset
diff --git a/mindspore/ccsrc/minddata/dataset/audio/kernels/frequency_masking_op.h b/mindspore/ccsrc/minddata/dataset/audio/kernels/frequency_masking_op.h
index 2da79d5261e..6bc48b6c51b 100644
--- a/mindspore/ccsrc/minddata/dataset/audio/kernels/frequency_masking_op.h
+++ b/mindspore/ccsrc/minddata/dataset/audio/kernels/frequency_masking_op.h
@@ -31,7 +31,7 @@ namespace dataset {
 class FrequencyMaskingOp : public TensorOp {
  public:
   explicit FrequencyMaskingOp(bool iid_masks = false, int32_t frequency_mask_param = 0, int32_t mask_start = 0,
-                              double mask_value_ = 0.0);
+                              float mask_value_ = 0.0);
 
   ~FrequencyMaskingOp() override = default;
 
@@ -43,7 +43,7 @@ class FrequencyMaskingOp : public TensorOp {
   bool iid_masks_;
   int32_t frequency_mask_param_;
   int32_t mask_start_;
-  double mask_value_;
+  float mask_value_;
   std::mt19937 rnd_;
 };
 }  // namespace dataset
diff --git a/mindspore/ccsrc/minddata/dataset/audio/kernels/time_masking_op.cc b/mindspore/ccsrc/minddata/dataset/audio/kernels/time_masking_op.cc
index 0bf309955ef..c93c1198e53 100644
--- a/mindspore/ccsrc/minddata/dataset/audio/kernels/time_masking_op.cc
+++ b/mindspore/ccsrc/minddata/dataset/audio/kernels/time_masking_op.cc
@@ -24,7 +24,7 @@ namespace mindspore {
 namespace dataset {
 
 // constructor
-TimeMaskingOp::TimeMaskingOp(bool iid_masks, int64_t time_mask_param, int64_t mask_start, double mask_value)
+TimeMaskingOp::TimeMaskingOp(bool iid_masks, int32_t time_mask_param, int32_t mask_start, float mask_value)
     : time_mask_param_(time_mask_param), mask_start_(mask_start), iid_masks_(iid_masks), mask_value_(mask_value) {
   rnd_.seed(GetSeed());
 }
@@ -35,8 +35,9 @@ Status TimeMaskingOp::Compute(const std::shared_ptr<Tensor> &input, std::shared_
   // input <..., freq, time>
   CHECK_FAIL_RETURN_UNEXPECTED(input->Rank() >= 2, "TimeMasking: input tensor is not in shape of <..., freq, time>.");
   TensorShape input_shape = input->shape();
-  CHECK_FAIL_RETURN_UNEXPECTED(input_shape[-1] >= time_mask_param_,
-                               "TimeMasking: time_mask_param should be less than the length of time dimension.");
+  CHECK_FAIL_RETURN_UNEXPECTED(
+    input_shape[-1] >= time_mask_param_,
+    "TimeMasking: time_mask_param should be less than or equal to the length of time dimension.");
 
   std::shared_ptr<Tensor> input_tensor;
   // typecast
diff --git a/mindspore/ccsrc/minddata/dataset/audio/kernels/time_masking_op.h b/mindspore/ccsrc/minddata/dataset/audio/kernels/time_masking_op.h
index cf2ef07287e..7099a2cbb27 100644
--- a/mindspore/ccsrc/minddata/dataset/audio/kernels/time_masking_op.h
+++ b/mindspore/ccsrc/minddata/dataset/audio/kernels/time_masking_op.h
@@ -30,8 +30,8 @@ namespace dataset {
 
 class TimeMaskingOp : public TensorOp {
  public:
-  explicit TimeMaskingOp(bool iid_masks = false, int64_t time_mask_param = 0, int64_t mask_start = 0,
-                         double mask_value_ = 0.0);
+  explicit TimeMaskingOp(bool iid_masks = false, int32_t time_mask_param = 0, int32_t mask_start = 0,
+                         float mask_value_ = 0.0);
 
   ~TimeMaskingOp() override = default;
 
@@ -41,9 +41,9 @@ class TimeMaskingOp : public TensorOp {
 
  private:
   bool iid_masks_;
-  int64_t time_mask_param_;
-  int64_t mask_start_;
-  double mask_value_;
+  int32_t time_mask_param_;
+  int32_t mask_start_;
+  float mask_value_;
   std::mt19937 rnd_;
 };
 }  // namespace dataset
diff --git a/mindspore/ccsrc/minddata/dataset/include/dataset/audio.h b/mindspore/ccsrc/minddata/dataset/include/dataset/audio.h
index b2fa960ad20..2bb890e0f28 100644
--- a/mindspore/ccsrc/minddata/dataset/include/dataset/audio.h
+++ b/mindspore/ccsrc/minddata/dataset/include/dataset/audio.h
@@ -215,13 +215,13 @@ class FrequencyMasking final : public TensorTransform {
  public:
   /// \brief Constructor.
   /// \param[in] iid_masks Whether to apply different masks to each example.
-  /// \param[in] frequency_mask_param Maximum possible length of the mask.
+  /// \param[in] frequency_mask_param Maximum possible length of the mask, range: [0, freq_length] (Default: 0).
   ///     Indices uniformly sampled from [0, frequency_mask_param].
   ///     Mask width when iid_masks=true.
-  /// \param[in] mask_start Mask start when iid_masks=true.
+  /// \param[in] mask_start Mask start when iid_masks=true, range: [0, freq_length-frequency_mask_param] (Default: 0).
   /// \param[in] mask_value Mask value.
   explicit FrequencyMasking(bool iid_masks = false, int32_t frequency_mask_param = 0, int32_t mask_start = 0,
-                            double mask_value = 0.0);
+                            float mask_value = 0.0);
 
   /// \brief Destructor.
   ~FrequencyMasking() = default;
@@ -242,13 +242,13 @@ class TimeMasking final : public TensorTransform {
  public:
   /// \brief Constructor.
   /// \param[in] iid_masks Whether to apply different masks to each example.
-  /// \param[in] time_mask_param Maximum possible length of the mask.
+  /// \param[in] time_mask_param Maximum possible length of the mask, range: [0, time_length] (Default: 0).
   ///     Indices uniformly sampled from [0, time_mask_param].
   ///     Mask width when iid_masks=true.
-  /// \param[in] mask_start Mask start when iid_masks=true.
+  /// \param[in] mask_start Mask start when iid_masks=true, range: [0, time_length-time_mask_param] (Default: 0).
   /// \param[in] mask_value Mask value.
-  explicit TimeMasking(bool iid_masks = false, int64_t time_mask_param = 0, int64_t mask_start = 0,
-                       double mask_value = 0.0);
+  explicit TimeMasking(bool iid_masks = false, int32_t time_mask_param = 0, int32_t mask_start = 0,
+                       float mask_value = 0.0);
 
   /// \brief Destructor.
   ~TimeMasking() = default;
diff --git a/mindspore/dataset/audio/transforms.py b/mindspore/dataset/audio/transforms.py
index 0bbc0191b08..a56bd830161 100644
--- a/mindspore/dataset/audio/transforms.py
+++ b/mindspore/dataset/audio/transforms.py
@@ -273,9 +273,10 @@ class FrequencyMasking(AudioTensorOperation):
 
     Args:
         iid_masks (bool, optional): Whether to apply different masks to each example (default=false).
-        frequency_mask_param (int): Maximum possible length of the mask (default=0).
+        frequency_mask_param (int): Maximum possible length of the mask, range: [0, freq_length] (default=0).
             Indices uniformly sampled from [0, frequency_mask_param].
-        mask_start (int): Mask start when iid_masks=true (default=0).
+        mask_start (int): Mask start takes effect when iid_masks=true,
+            range: [0, freq_length-frequency_mask_param] (default=0).
         mask_value (double): Mask value (default=0.0).
 
     Examples:
@@ -304,9 +305,10 @@ class TimeMasking(AudioTensorOperation):
 
     Args:
         iid_masks (bool, optional): Whether to apply different masks to each example (default=false).
-        time_mask_param (int): Maximum possible length of the mask (default=0).
+        time_mask_param (int): Maximum possible length of the mask, range: [0, time_length] (default=0).
             Indices uniformly sampled from [0, time_mask_param].
-        mask_start (int): Mask start takes effect when iid_masks=true (default=0).
+        mask_start (int): Mask start takes effect when iid_masks=true,
+            range: [0, time_length-time_mask_param] (default=0).
         mask_value (double): Mask value (default=0.0).
 
     Examples:
diff --git a/tests/ut/python/dataset/test_frequency_masking.py b/tests/ut/python/dataset/test_frequency_masking.py
index 3893b2f13bb..fc77bca5c69 100644
--- a/tests/ut/python/dataset/test_frequency_masking.py
+++ b/tests/ut/python/dataset/test_frequency_masking.py
@@ -119,7 +119,8 @@ def test_frequency_masking_invalid_input():
     test_invalid_input("invalid mask_start", False, 2, 100, RuntimeError,
                        "MaskAlongAxis: mask_start should be less than the length of chosen dimension.")
     test_invalid_input("invalid mask_width", False, 200, 2, RuntimeError,
-                       "FrequencyMasking: frequency_mask_param should be less than the length of frequency dimension.")
+                       "FrequencyMasking: frequency_mask_param should be less than or equal to the length of " +
+                       "frequency dimension.")
 
 
 if __name__ == "__main__":
diff --git a/tests/ut/python/dataset/test_time_masking.py b/tests/ut/python/dataset/test_time_masking.py
index 3aee6340590..1ef77b3eec8 100644
--- a/tests/ut/python/dataset/test_time_masking.py
+++ b/tests/ut/python/dataset/test_time_masking.py
@@ -119,7 +119,7 @@ def test_time_masking_invalid_input():
     test_invalid_input("invalid mask_start", False, 2, 100, RuntimeError,
                        "MaskAlongAxis: mask_start should be less than the length of chosen dimension.")
     test_invalid_input("invalid mask_width", False, 200, 2, RuntimeError,
-                       "TimeMasking: time_mask_param should be less than the length of time dimension.")
+                       "TimeMasking: time_mask_param should be less than or equal to the length of time dimension.")
 
 
 if __name__ == "__main__":